YouTube Transkripti:
IOT Data Hackathon 2026 – Workshop 3

Videoları baştan sona izlemeye gerek yok — tam transkripti al, anahtar kelimeleri ara ve tek tıkla kopyala.

Paylaş:

AutoDub

YouTube'daki Yabancı Videoları Anla

YouTube'u Türkçe Seslendirmeyle İzle

Dil engellerini aş, dünyanın dört bir yanındaki kaliteli içeriklerin keyfini çıkar

Ücretsiz Kullan

Video Transkripti

Video Özeti

Summary

Core Theme

This workshop explores the evolution and application of AI in automation, from intelligent automation and AI agents to AIoT, emphasizing practical implementation and the transition from prototype to production-grade solutions.

Mind Map

Genişletmek için tıkla

Tam etkileşimli Mind Map'i keşfetmek için tıkla

Hello everyone and welcome to workshop

3. Uh it's very great to see you all

again. Let's get settled as we wait for

a few more friends to join us today. Before

Before

we start, um, let me do a quick audio

check. Can everyone hear me clearly? If

Thank you. Great.

So in our last section on Tuesday, we

covered data presentation and teaching

skills. Today we're moving on another

hot topic which is AI. This year um we

are introducing the best use of AI

award. So implementing AI will be a key

requirement for winning it. Today's

schedule is focused on the theme

empowering innovation from AI

foundations to agentic AI development.

I'm I'm your MC Selena and we're

delighted to have two expert speakers

join us to share their insights. So

during their present uh during their

sharing if you have any questions please

feel free to type them in the chat. Our

speakers will address them address them afterwards.

So for the first section we will have

from intelligent automation workflows to

We are pleased to have Mr. Edward Lao

chairman of innovation and

transformation committee Hong Kong

artificial intelligence industries

association. Hi Edward. Um could you

please share your presentation

PowerPoint with our participants?

>> Sure. Um, how are everyone?

>> Thank you.

>> Can you see my screen?

>> Yeah. Yeah, we can see it.

>> All right. Okay. Good. Hello, everyone.

Um, all right. So um it is my pleasure

to um share with all of you um and um

I'm Edward from the Hong Kong artificial

intelligence industry associations. So

um today I'm going to share with you

about a um triple AA so from intelligent

automations referrals to AI agents and

AIoT. So the key reps here or what I'm

going to share today is automations,

All right. So um I'm to um walk through

the different topics including um

intelligence automations to AI agents

and how both of it working togethers and

um using a retail case and and finally I

will share about AOT because this is IoT

hackform right um so give you some more

idea about um maybe another option you

can uh put in to your solutions. All

right, before I start, let me give um a

a minute about the association and

myself. Um actually um the association

actually we are focused on AI and it is

formed by a um group of business

leaders, entrepreneur and innovators and

we our mission is to drive innovations,

strength the industry and create a

long-term impact for our members as well

as the societies.

And uh we have a um honor honorary

advisors um to support the associations

from governments from legislative

council members, chairman of authorized

authorities um charable foundation, a

president president of a university in

Hong Kong and also the former secretary

of innovation and technology of um the government.

government.

All right. So um a u some brief about

myself um well I'm you know um of course

same as you many of you here in the

technology area um for a while and also

um I'm also a partner in the TDC council

um as a task force focus on um the GBA

um and also um supporting the um

Australia accountants using it or AI to

improve their work right and also Um

actually I'm also um wanting data beside

AI and data is very important right so

uh probably you will see me again in

another section about data all right so

and um you see the last one is um um I

I'm also appointed by the um WSA will

submit award and it is aligned with the

United SDGs

uh because the hackathon um one of the

theme is related to sustainability so um

SDG is quite important so um let me

highlight and you

one of the case I will highlight a case

that it is aligned well aligned with the

SDG of um United Nations. All right. So

um back to the topics. So automation or intelligence

intelligence

um many of you um is is quite familiar

with AI right? Um but

um actually a a lot of people u ask ask

me or the association about what's the

difference between automation and intelligence

intelligence and

and

we have AI then is of course is

intelligent right so so did the AI we

have the AI or AI agent do we still need

automations what do you think well um in

um after we we we we study a lot um with

the the um enterprise um or or the

operations well definitely um it work

togethers it is not replace one another.

So talking about the intelligent

automations one one key thing here is

RPA robotic process automation

um some of you may be quite familiar

with it because it is not a new thing

actually it is traditional automation

I'll say that RPA actually is already in

the market more than 10 years um however

it is still a very good tools it is very stable

stable

very efficient because it is it is

applied to a um fixed

referral for example from A to B. So um

if the refer is fixed and it will do it

automatically for 100 times thousands

time 7 by 24 so it's really efficient um

and and save a lot of time and cost and

um even though it is a traditional uh

tools but I would say that like airport

or even mtr a lot of larger enterprise

they still using uh they still have RPA

even though they definitely they have AI

I agent or AI um solutions right um so

it's different tools um so that's why um

I for for your solution I would still

recommend some of the workflow that it

is fixed and you need to repeat do it

again again and again for example you

have 10 people's doing the same thing um

every day for one hour so then is 10

people then will be 10 hours a day and

then for a month 100 hours and then you

save you a lot of uh network um and not

need to use an AI agent and the RPA's

work closely with the EP data and IoT

data or internet for example. A typical

example is maybe u capturing the I

capturing the data from the IoT device

or sensor or from the internet and then

input the data into your CR system or

ERP system to input instead of human

input and then they will do it a

keystroke. All right. And and after that

and you can maybe also send after the

the input into the ERP it will send an

email to the recipient and then they

will do it automatically. That's quite

when you hear the scenario is a little

bit similar to AI agent but it is if it

is a fixed path all right and then okay

is one of the the good choice so how

about intelligent automation intelligent

automation it is using AI to power the

RPA so which means it is AI powered RPA

for example um um well let me um using

um one of the um uh popular tools for

example Microsoft Power Automate or

Power Apps. Um, of course it is already

in the market for a while for many

years, right? Um and now with the with

the I mean with the power of AI then to

make the power automate actually power

automate it is a uh one is it is one

type of RPA first of all I'm sorry but

automate is one type of RPA and it can

connect with different source document database

database

excel specific shareepon and then it

connect to the power apps to different

people. All right. And then of course it

is quite use easy to design web flow.

Right. So just um drag and drop to

create workflow. So uh actually it's

already in the market already for many

years right. This is this is very easy

to use it and

and

with the AI then it make it much more

easier. For example, instead of kick and

um just kick the refle

just using just like um copilot, right?

Or other lm just type it using a natural

language, right? So um create a

sharepoint is for example you can see

create share is send a email uh based on

different condition

and then um send a team notifications

and then get a approval process. So just

tell just just a long mode LLM right

just like um copilot. So you see a

copilot studio here. So input the um request

request

and then they will make the workflow for

you. So it is like a AI assistant for

the RPA. So the RPA it is a robotic

process automation. So automate the

workflow on a fixed path and with AI

power and then can even make the

workflow builder much more easy, right?

And also connect and then it will help

you to connect to all the data connection.

All right. So um um so just give you a

um um um quick idea about um

intelligence automation. So um

just like uh a lot of tools uh for

example YP or CRM that using AI to make

the tools much more usable and and my

suggestion is uh of course we have a lot

of u innovative technologies right um uh

but however it doesn't mean um the

traditional tools doesn't work or have

no value the good thing from my point of

view from a practical point of

um using some traditional stable and

effective productive tools is still

good. So what I'm going to do is using

AI to empower those traditional tools as

I mentioned using AI to to power for

example imagine if there's a company

your company already have a lot of

traditional tools like RPA is quite

popular and and um before that RPA uh

because it only apply to fix workflow

um before that only the IT teams um have

a capability

to confect the workflow once your up to

to to conf I'm sorry once your workflow

has changed but now with um AI so the

user can use it's quite easy to config

the the the RPA workflow or create even

create new RPA workflows

um so AI plus um it is what something I

suggest u um um not only in the

reliability and also in your solution

because um uh we have to to to balance

right to make your solution much more

sophisticated. All right. So um AI agent

um all right so

first of all um I believe many of you is

familiar with a agent right because uh

even you didn't use it before or you

didn't create build your agent before

definitely you heard about AI agent and

and um you have idea about agent um

today uh because time is limited um I'm

not going to demonstrate how to build a

AI agent create because it's easy. Uh I

mean it's it's a there's a lot of

resources. I mean you can find hundred

hundreds of um um videos on the YouTubes

showing how to do a AI agent and also um

I believe um the next speaker from AWS

um we will do it a lot. I mean to to

demonstrate about AI agent. So um but on

the other side I I'll share with you

about um some practical tips about when

you doing the AI agent and also how to

using AI agent working together with RPA

or intelligence automations to together

um um to create a solutions and also how

we apply to the AIT right so um a agent

um I mean a basic um knowledge is to

reasons learn uh and adapt to the

dynamic environment and make decision.

So the one big difference between AI

agent and AI and RPA it is RPA it is on

a fixed workflow. Okay from A to B or

ABC all right fix um if the scenario is

different environment is different it is

not apply to RPA and then AI agent is

autonomics right then you can work on

this because it has a reasoning behind

and also he can make decision uh make

the decision make um um um

action that suggest the action that is

which this RPA um doesn't have. All

right. So that's why um I mentioned here

a agent is enhanced automation by adding

intelligence to automation but still

rely on automation tools to execute the

back end structure task but not

replacing RPA. For example, even data,

every company you have

structured data and unstructured data,

right? Well, we see a lot of company

they they didn't um well adopt AI

because they don't have enough AI ready

data or AI ready referral.

So one thing is using AI using AI

to transform the unstructured data to a

structured data.

All right. and then to leverage other AI

tools to do the automation or

intelligence work right so um just two

reflection question here so uh first one

is so u when you deploy um AI agent so

which of the current workflow in the in

the system um in operation is rely on

strike rules fix rules and which require

judgment so it is if on the strike rules fix

fix

RPA probably is a good choice

automations but however it require

judgment then it need to be um using AI

agent right so to to analysis the

workflow first

and think about it if we combine AI

agent together with RPA then uh what is

some new opportunities in the process

that we can discover

right so um think about this

and also when you when you um deploy

your AI agent or design AI agent. Um

also think about the AAA um

to see whether it is fully optimate or

it require human in the loop. So for the

workflows okay or on the on the

automation process okay we have to

understand which one um it is automated

and which one we require human in the

loop. All right. For example,

um customer service

um when of course uh there's a lot of

use case using AI agent um or for the

customer service um chatbot, right? But

however to handle com customer complaint

imagine if you're the customers you

already pissed off um I I don't think um

um have to have the the the machine you

don't want to talk to machine right

because because you have to campaign you

don't want to talk to machine you you

want what do you want at least you have

someone some customer um customer

service people from the from the company

right to to understand to help Right. So

and it is a typical case that um human

need to tribute

and then um about authorities. So so um

in the workflow when you design the AI

agents so what's what the system or or

device that the AI agent is need is

allowed to control or limit all right

and and then accountability. So um AI

agents are powerful can do a lot of

things automatically right and

intelligently but we we should have

approval audit audit trials safety and a

risk control and it can have a audit

trial we can uh chase back right so

something um need to consider it

um about IoT then uh let me also um

touch on the supply chain optimization s

so um for the IoT or AI agents and it of

course it may need to con connect to the

ERP or inventory system. So it will

capture data on the inventory levels or

supply need time or transportation costs

right. So the AI model will doing the

predict demands recommend to reorder

points or safety stock

and pick a purchase order. So it will

have a tot bunch of um action sequential

actions based on the AI agent and even

um one thing I think um I would see a

lot of um company um using AI is doing

um what if analysis for example what hap

what happen if the demand rise by 20%

then um how's the supply chain how's the

inventory the delivery inventory

delivery Okay. And then uh what if this

one of the supplier fail right we'll see

a lot of case that uh one of the key

supplier um the fail um especially

during imagine co one of the supplier

fail and the whole supply chain is

broken and then you lost the delivery.

All right.

>> All right. So, um let me um uh um also

uh run through a um video is about um on

AI agents about the quality variation.

So, it is based on um customer service.

So there's a there's a chatbot or or

customer service to handle

to handle customer uh inquiry and how to

use it um to improve the quality eration.

>> The quality evaluation

agent analyzes completed conversations

and cases

assessing them at scale against our

organization's quality standards. I see

we're hitting our targets. Though the

agent is alerting me here to a low

score, I can drill into the evaluation

summary and review proactive suggestions

on how to improve with coaching recommendations.

So in in that um

I'm sorry I think [clears throat]

oops I'm stopped.

>> The quality evaluation agent analyzes

completed conversations and cases

assessing them at scale against our

organization's quality standards. I see

we're hitting our targets. Though the

agent is alerting me here to a low

score. I can drill into the evaluation

summary and review proactive suggestions

on how to improve with coaching recommendations.

recommendations.

In this case, adjusting the empathy and

tone and refining processes executed by

the selfservice agent when resolving the

issue going forward. Now the next time a

customer reaches out, our Aentic contact

center powered by Dynamics 365 is ready.

Intelligently evolving with the latest

intents, up-to-date knowledge, content,

and essential quality guidance to help

resolve issues and build.

or in this um demonstration you'll see

that um there's a lot of um customer

case and then um the agents to analysis

um the conversation between the

customers and the customers uh

representative um either verbally or on

the chat on the chat box. So um and

Daniel analysis to and then um to give

suggestion how to improve it and for

each of the case to give a overall score

and to keep improve the the customer

service. So it is one of the case um

there's a um other case um you can find

on the internet so just give the case

that um you can take a look so uh about

AI agents.

Um so as I mentioned before um

in [clears throat] intelligent

automation or RPA or AI agent they all

they have their own advantage right um

one is cost effective and the other one

is automous and intelligence um it work

well together and not replacing one

another. So um in um in the there's a

real case that I would like to share

with you is um demonstrate how

those two um air tools working together.

All right to compete um um the scenario

and for example this case um is a

promotion click to order. So for example

there's a retailer there's a brand

retailer they have um um different kind

of promotion on the social media okay

and of course they can using um um AI so

using AI as AI agent um as the marketing

agents as a shopping assistance as a

chat box customer service and when the

customers receive the um the promotion

on the internet and if feel interest

kick it right click it And then maybe

there's a chat. He will uh he or she

will be also browse the the internet uh

browse the the website of the company to

see um whether the product he's

interested and then um he he may also um

using a chatbot to discuss for example

to um to ask about the features function

feature about the products and also

place the order on the internet right on

order. So we have different role AI

agent as a marketing um agent

um shopping assistant chatbot and also

um RPA and RPA will be a operational um

execution doing operation executions

um for example doing the order entry and

also update the system data

and one thing is very important is human staff

staff

All right. Um to handle some exceptional

case or campaign as I mentioned also we

So um here's the scenario for example on

the stage one. So we have a customers AI

agent RPA and human style. So for the

stage one there's a social promotion

um and the customers will will see a

social media on the pro uh um he he or

she will click the promotion on the

social media. All right, once you click

it and then the AI agent will do

something actually before that of course

the AI agent the marketing team can

using AI agent to generate a promotion

to a target uh customers and once the

customer click on the social media and

agent can do a personalized NBA next

best action. All right. And in the back

end, the RPA can help to check uh doing

the tracking of the performance of the

um the promotion to cake or to do

analysis or updating the the

advertisement platform or CRM for human

stuff for the marketing definitely he

can um reveal the whole process for

example for promotion to kick

performance to see how well the social

media is doing well. For the stage two

when the customers continue to browse

the products to view the details of the

products function features and then the

AI agent

uh will help to make a scoring about the

interest how interest the customer will

be from high medium to low in a

real-time basis and determine the

segment of the customer's interest and

then for the back end the RPA will keep

maintaining the the customer portfolio

and and the behavior locks um into the

CRM. Uh and

and

um the same concept for human staff, he

will reveal the clickthrough and crow performance.

For stage three,

um for example, for the customer, he

view it, but he didn't buy it. He didn't

put in the shopping cart. And then um

the what the AI agent do maybe he will

send a reminder or some educational

content or some discount offer to the

customers. So the customer will receive

more retargeting

uh advertisement with the personalized

offer. So it is what the AI agent can

can do. And for the RPA very simple

update the CRM and then um and trigger

some u sequ sequential actions. And for

the human staff can we will per review

And for stage four,

the customer maybe the customer feel

interest because um he once again

receive um some special promotion um

from from the AI agent and then um he

would like to check for example the size

of this um of the curve or the delivery

terms and then he um communicate with

the chatbot that will the AI agent will

handle um answer the customer's requests

and give recommendations

and for RPA Okay. And we'll be in the

back end we update the contacts and the

logs in CRM

and also um the high intent for the

customer follow up.

All right. And for the human staff and

then um if this if the RPA u make it as

a high intent for the human fault and

then the the human staff can trip in and

to handle the to lead the conversation

with the customers. Okay. It depends the

product. For example, some um fashion

garment maybe not necessary but some for

some um luxury products um or

professional service um um it is mostly

happens. All right. So, so the AI agent

and the RPA work together to uh measure

measure the the need from a um from a

potential to high potential uh prospect

and then they and then the human staff

will be should jump in to um handle have

a more human touch um conversation or

And then for the next stage

um maybe the customers u maybe feel

interest to continue the dialogue or

maybe the customers u keep silence

didn't respondse right and and and and

maybe he's cool down and then AI agent

will doing the MBA the next action maybe

uh based on his analysis of the customer

behavior and then maybe he send more um

educational content um special offer or

we assign it to the client, okay, based

on um the um personalized behavior and

the RPA will um keep updating the status

or even escalate to a human task,

right? And then for for the human staff,

maybe he maybe he his job is need to

follow um with the customer by himself

And finally the customers confirm the

purchase and then they need to um uh key

in the delivery information and the

payment. And here um AI agent will

become a shopping assistance to um have

a guidance to the customers to check out

to fill in the datas. All right. And

you'll see that for for the AI agent

actually it is um at the beginning the

AI agent is a marketing assistance to

create a u the promotion contact and

then distribute to um the social media

and then in the in the middle of the

workflow and then AI agent we have AI

agent as the chatbot okay to answer the

quiry from the customers and when the

customers confirming the purchase and

then is becoming another sales uh

assistant or shopping assistant to help

the customers to um to check out to

finish the check out the whole process.

So um so AI agent actually um is quite

powerful here and do a lot of different

jobs. Um but on the other side RPA still

very important and is doing the back end

the workflow right so so for example

keep updating the the CRM based on the

customer's uh behavior and also to uh

also update the the the performance or

the doing the analysis about um the advertisement

platform um to the internal CRM

or even once the order is confirmed is

creating orders or doing the process

payment. So all um for example the

process payment or um confirming the

delivery and logistic is quite routine

or fixed referral that is perfectly good

for the RPA

right so for the for the human uh being

the order or even maybe handling

complaint for example the customer order

uh make the order but it didn't deliver

on time or it delivered to a wrong

address and the customers have a inquiry

or even campaign. So, and now um the

customer may not want to talk to a

machine, right? And then um there's a

human stuff um maybe jump in and then to

handle it. Um so you'll see that um for

the AI agent we can use in different

touch point right from the marketing to

the chatbot um to the shopping

assistance and RPA doing the operation

automation and the human we

handle um specific task exceptional task

like campaign. So which is um is how all

those three parties working to together

to serve the customers.

All right.

Um so for the for the retail case um

hopefully can give all of you um an idea

about um using using um the RPA together

with AI agent um to make um to make the

customers journey or the c have a more

have a better customer experience right

and we more not just automate intelligence

intelligence

um and also um will be more effective,

right? And uh on one side it is increase

the um effectiveness of productivity and

on the other side is also uh increasing

the customer's um experience, right? Um

so think about that. So um for the last

topic um it is um AIoT

um well for the for the hackathon it is

about IoT right internet of thing uh but

um how but today I will also like to uh

share with you about um AI IoT um so you

can also think about it this is maybe

another alternative that um can you can

um um improve your your solution right

um to handle those challenges.

So um I'm not sure how many of you is um

familiar with AIoT.

All right. So AIOT it is a integration

of AI with IoT to create smarter more

autonomous system that can sense

analysis data and make decisions. All

right. So imagine I of course it we

collect a lot of data from the device

right from the sensor

and then sending to the crowd and doing

analysis or um um um after actions but

with AI it help to not just sense the

data and ask the data and make decisions

is how AI is make it more intelligence

right so AI agent and even RPA can using

AI data to trigger action s across the

system and device. So this is how we we

collaborate together. So AI IoT device

the device or the sensor will become the

eyes and ears right to capture real

world data that um we all know that so

and AI is as the brain to analyze the

um a similar actually it can also apply

uh um um to a customer service that um I

will share with you later on uh but of

course uh AI IoT it is u there's a lot

of um ao applications um in the in the

So think about that um um you have a lot

of different device and and when you

design the device and also think about

how we can using AI

to um analysis the data. Of course, not

all the data that um you have to grow

for the AI, but to think it how we using

AI to create a um more business value as

as well as the um impact, right?

And um when you when you um if you you

are interest um to build a ao in your solution

solution

um think about it. So when you design

ALT um think about the device

connectivity edge layer data platform

air service and application right

because it's um different from IoT uh

AoT um using edge compute right um um

for IoT of course we have the sensor and

then you send to the cloud right but for

AoT we have um edge layer to handle the

data and then um we'll be So it improve

the um latency um connectivities and

even security because not all the data

is going to the cloud. it just do do all

the work and then it will send us the

the selective data um back um to the to

the headquarters to the cloud right um

and for the for the hacker requirements

and think about um how does your AIT

framework can map to the requirements on

um a simple comp um table compare IoT

and ALT. So um for example um for the

functionality IT is so both is working

on connectivity. Uh but for IoT it is

connect with data and doing monitoring

and remote control. For IoT

the connectivity is work together with

intelligence analysis and prediction and

uh automate action. prediction is is one

of the um popular use case in AoT. All

right. And for the intelligence level of

course ALT is more ALT is uh have a

higher intelligence level including

machine learning um pattern

recognizations and continuity that's why

it have prediction

for data processing and transmission

right so for IoT the raw data is sent to

the cloud so um when the sensor capture

data and then you send the raw data to

the cloud and and do a further u data

processing. So um all the raw data will

be transmit um

to to to the but for IoT is a little bit

different. So there's a edge computer

edge computing using edge computing the

framework. So we um processing on the

edge device and then sending the um

um the related um data inside your

events um to the cow. So not all the

data use okay we will store in the in

the edge computer for latency and

response. So um IT definitely has a high

response um because it depends on the

network and the cloud right

and but for ALT

is more real time because uh most of the

thing is um done in the edge computer.

All right. Um and some typical use case

it is um for example status monitoring

um alert remote control for AIT is

predictive maintenance

um anommy detections to prevent um um um

um authorize people assets right and

doing a vision analytic or computer visions

visions

all right um for here I would like to

share um a case um there is uh related

to the SDG sustainable development

goals. Um actually by um the framework

under United Nations there's um 17 SDG

um that uh because this is it is um also

maybe um somehow related to the um GS1

uh data hack about sustainability but of

course not all 17 SDGs related to that.

um you have to pick some of them. For

example, number nine um is um uh um

about infrastructure and number 11 it is

about smart um urbanization or smart

city thing. Um and and just let me

introduce this um WSA is world submit

award. World submit award. It is a

global global um competition global

competition uh across 182 countries in Asia

Asia

um um and Europe um and even Africa. So

um Hong Kong, Hong Kong, China, Japan,

Korean, um Indonesia, from Indonesia to

Australia. Um every countries every

years we have um uh come um um startup

programs to join the competition um just

like um JS1 here. And uh actually I'm

honored to um uh as one of the national

expert in Hong Kong um to nominate the

um the project and and actually next

week just because yesterday I I have I

also have a meeting uh because next week

um um because I'm doing the um the all

night um panel

judge uh across the the first one about

180 countries and then I'm the f the

first one judge and then the the next

one will be um in Brazil just next week.

Um I just have a meeting yesterday to to

to have one of our um because there are

two company um this year we have two uh

Hong Kong local company. um they we we

we are we are select and then we also

win um the first the first S list and

then now they go into the second S list

and then they will do the pitching uh

next week uh on on Brazil and and I'm

working with them um and I am and I will

pick one of them one of the one of the

case that we will be uh present uh next

week in Brazil um they are using on AI

IoT so that's Why today I think it's

good to share with all of you. Um so

it's a Hong local Hong Kong AIT

solution. Um they're working on uh

different scenario from smart cities to

crowd managements, logistic uh property,

right? So um the using the AoT they do a

lot of different thing for example from

the people counting

um to to um off um to to monitor

unauthorized um assets

um motion detection on the hospital um

speeding detection on or prevent the

parallel park um on the street um in

Hong Kong and also fight um smoke

detection in um different place in Hong

Kong, right? Um actually so it is it is

um the product so it is that they have

patrol um or or um

um the the monitoring so it's powered by

the edge AI so so the edge will be

actually will be here or in the inside

the pet show. So traditionally um we

have CV right for example for example in

the control room there's a security and

he have using his own eyeball

to monitor like 10 15 uh CCTV monitor

right uh but but u in reality it is not

possible and it's not um it's not really

effective because for example um his

mobile phone ringing or he um or he got

a WhatsApp message and then he look at

the mobile phone and then something

happened on the on the one of the CCTV

and he didn't catch right but um using

AI we we can have like 2000 or more um

CCTV of course it is is um not CC so

this is AI powered um um camera and then

it will doing in the back end we have um

using um using AIT to analysis um the

movement and then it give alert instant alert

alert

um to the security guy or even have

using um um some automated actions.

All right. So, so actually it is um some

of the case happen in Hong Kong. Um for

example, for this one, it is to um

help helping the the the city to find

out um those have a parallel park or or

u a legal parking, right? and even um

working with the traffic light and also

here in um it is the central and also

another one in uh um kite kaid right so

uh because the railway we have people uh

walking running or um bicycle and then

they doing the speed detection to and to

give a um signal um to to

um to alert the the bicycle to to slow

down. All right. And of course, it have

a BB BB to to the alert and also um

there's different kind of signal to um

to alert um the bicycle to um um to slow

down otherwise to to prevent some

accident with the the the people.

Another thing is um in the airport um

um to look at the career bell um in the

airport to see the the luggage um and uh

also on the

on the on the on the trust the trust

cargo for example this one you see that

you uh look at this it is

oops okay so for example here you check

it this is this empty here right or

something it is um the or or the puzzle

didn't put it um

correctly. All right. So um and actually

it's just some more use case um for

example for the crown management it is

um in the u Hong Kong exhibition center

uh because a lot of people right to line

up um or even um and in in other city uh

for example KL uh because imagine for

all urban city right uh Hong Kong Tokyo

right there's a lot of people um um car

management even crime actually um

traffic it is um is some of the issue

that all the well-developed country need

to be solved. So it is um SDG related

and actually um for for them it is they

also helping to um monitor the the

monkey in Hong Kong, right? Um I don't

have I I didn't put the video here. So

they they prevent they prevent the

people to doing uh um monkey feeding to

um to feed um to give the food to the

monkey because it's not allowed right

and also they um and and the the same

project is also roll out to Australia,

Melbourne um um or South Australia um

state and even Singapore because this

interesting because Hong Kong we have a

lot of monkey right so as I mentioned

they're using that to monitor prevent

people doing the monkey feeding and then

now um Singapore they still have um the

same issue then it's much more worse the

monkey even the monkey even um going

into the house and then now um they also

using the same system um from Hong Kong

to help them um um to doing um um um the

the p um the alert right all right so um

as I mentioned think about that um um if

your your solution it is relate you want

to related to the sustainability think

about that because um uh for this

project this is two SDG is related one

is number nine this is industry

innovation and infrastructure and the

other one is they have a uh sustainable

cities and um community or because um

smart urbanization um sometimes as we

call it 11. So um as I mentioned for all

welldeveloped country traffic

uh people car management crime

management crime um it is something we

need to solve. So um I'm think about it. All right.

So um also give you when you design your

your your solution um some of the

refreshing question I would um suggest

you to think about that for example for

the AoT stack um how does your AIT

framework map to the hyperform

requirements using the IoT data for

example uh equipment failures unauorized

assets uh or or the products the product

ID when you send a product um scan the

product ID and also um relate to the

open data or alternative data for

example weather, traffic, demographic.

So, so look at the challenge you

receive. Um, how how do you want the A

IoT framework to map to the requirement

using a IoT datas, open data and

alternative datas, right? And for the

for the sensor for the device um as I

mentioned IoT is the eye and the ear

right to collect the real world data and

once the data collect um if you're using

AI agent how do you want the AI agent to

handle the data

all right and the frequency um I mean to

to do it continuously every minute or

once a day it depends on the scenario

right so and what the AI agent design

because um it is intelligent right um um

how and when it escalate to the human I

think is is it important for example um

for the customer service for example if

the queue if the queue length is more

than six customers and they're waiting

here more than three three minutes right

and and uh people get u impatient right

then um you need to open a do additional

check out if staff available right so

it's it's quite make sense then how when

it applied to the AI agent uh what it

need to decide and what to do so for

example to detect using the

simple example here so uh to detect

there's a long queue more than six

customers on the queue and more than

three minutes and then suggest uh

additional child and then to confirm

this staff availability and then um open

a extra check out with um um the digital

sign uh signage to inform the customer

hey there's another one right so it is

simple use of the um um AI IOT

all right and um finally um one thing we

have to when when you design your

solution think about the AI IoT how to

turn AoT data into insight and business impact

impact

For example, um when you receive your

challenge, right, you have to to think

what is your IoT datas, maybe

temperature, uh p movement, location or

specific port ID. So it's your IoT data,

right? Um and then plus your open or

alternative datas, weather, traffic, um

or even ESG data set, right?

And then they go into the AI mode, go in

the AI model to analysis the pattern,

generate predictions, give you insight,

right? Um the the key thing here is

using the AI power to doing um some to

give um uh valuable insight for example

prediction which of course is aligned to

the hyperon stream or the challenge,

and then convert the insight into a

business value and impact. This is the

the the key things.

For example, for example, uh preventive

maintenance and safety. Um for example,

uh for safety, for example, this old

building or this theme park, right?

Whoops. And and how using um the IoT

datas and to um um to to doing uh

preventive maintenance on the machine.

For example, it is a old building or

transportation, right? for

transportation and it needed punitive uh

maintenance to uh provide safety for the

human for the human rights. So is

sustainability or second one supply

chain supply chain optimization or

demand forecast then maybe you can using

some open um data like weather traffic

right with with um with the your own IoT

datas um to optimize the supply chain or

doing um for example if one of the

supplier um stop right or the demand

increase or for the retail case uh

sometimes uh rain um I mean sunshine day

or rain day uh is definitely affect the

the affect the the retail right so it's

doing um demand forecasting and also

doing the people contract in the in the

um um in the shop and also how it can

provide better customer service while

improve resources efficiency and

sustainability that is one thing that ao

is doing. So as I mentioned before, we

have to um provide better customer

service, better let them have a better

experience. But on the other side, we we

didn't instead of putting more resources

to increase the cost, we have to using

technology right innovations to to

improve the resources with uh efficiency

and even make it a sustainability

uh advantage. That is something um you

can think about that um in in in your

solution when you um to um to tackle the

the the challenge, right? Um so so um

think about that. So um this is pretty

much that um so actually this is pretty

much that um I have for today. So we

have walked through the intelligence, we

have worked through the um intelligence

automations including RPA you using

using a um traditional traditional tools

uh with AI power can still um deliver a

lot a lot right and we have AI agents

of course AI agent is is is is really

good but also concerned about the

governance right u um the risk security

when we have human in the group, we have

human involved when you're using uh AI

agent. This is something I would like to

mention. Then how can we using um um AI

agent work together with the RPA to or to

one plus one more than two, right? Uh

how we can using that and plus AI agent.

Um um think about the retail case,

right? We have AI agent, we have RPA and

also we have the human right and when

doing the customer complaint we have the

human and the AI agent do a lot doing

the customer um the marketing assistant

to a chatbot customer service and

shopping assistance right and for a IoT um

helping create more much more business

value and think about also sustainability

sustainability

U that it is um I think this is what

innovation can can help right all right

so um I think it's pretty much about um

for today um tripleA so um thank you

very much and good luck u on your hackathon all right and uh thank you

hackathon all right and uh thank you very much

very much >> thank you Edward for your detailed um

>> thank you Edward for your detailed um sharing and explanation on the words um

sharing and explanation on the words um so just now Um I saw there are a couple

so just now Um I saw there are a couple questions from the floor but due to the

questions from the floor but due to the time limitation we'll leave it to the

time limitation we'll leave it to the Q&A section and answer them later on. So

Q&A section and answer them later on. So right now we'll move on to the second

right now we'll move on to the second section first.

section first. So for the sec Thank you Edward again

So for the sec Thank you Edward again for the second f uh second section it

for the second f uh second section it will be about kirao agentic AI

will be about kirao agentic AI development from prototype to

development from prototype to production. Um we are pleased to have

production. Um we are pleased to have Mr. Lo Young, partner solutions

Mr. Lo Young, partner solutions architect from Amazon Web Services to

architect from Amazon Web Services to share with us. Hi Lo, could you please

share with us. Hi Lo, could you please share your PowerPoint with us?

share your PowerPoint with us? >> Oh yeah, of course. Um, let me share my

>> Oh yeah, of course. Um, let me share my screen.

screen. Thanks.

Okay, can you see my screen right now? >> Yes, perfectly.

>> Yes, perfectly. >> Okay. Yep. Perfect. Um, okay. Good

>> Okay. Yep. Perfect. Um, okay. Good afternoon everyone. Uh this is lock. Uh

afternoon everyone. Uh this is lock. Uh I'm the partner solution architect from

I'm the partner solution architect from Amazon web services and today I I set

Amazon web services and today I I set the topics about Kirro which is our

the topics about Kirro which is our agentic uh AI IDE and specifically we

agentic uh AI IDE and specifically we would like to talk about uh how to

would like to talk about uh how to leverage these services in order to help

leverage these services in order to help you to turn your prototype or develop

you to turn your prototype or develop your prototype and then subsequently uh

your prototype and then subsequently uh to a production grade products. But

to a production grade products. But before that I would like to start with

before that I would like to start with uh IoT specific topics like because uh

uh IoT specific topics like because uh for JZ there are specific topics on IoT

for JZ there are specific topics on IoT and I would like to share some of our

and I would like to share some of our customer cases and also use cases uh

customer cases and also use cases uh before I jump into how curo can help on

before I jump into how curo can help on these sections.

these sections. Okay. So here are the high high level

Okay. So here are the high high level agenda for today's sessions. So I would

agenda for today's sessions. So I would like to start with talking about some

like to start with talking about some IoT in AWS or even for AWS customers and

IoT in AWS or even for AWS customers and the second part will be how the AI is

the second part will be how the AI is changing and how software is built and

changing and how software is built and operated. And then the third one we we

operated. And then the third one we we will briefly go through Kirro IDE and

will briefly go through Kirro IDE and also there's specifically a uh concept

also there's specifically a uh concept or the features we embedded into Kira ID

or the features we embedded into Kira ID is about specdriven development how does

is about specdriven development how does it work and then finally we will talk

it work and then finally we will talk about some of the tips and tricks to how

about some of the tips and tricks to how to turn your prototype into a production

to turn your prototype into a production products. So let's get started.

products. So let's get started. Okay. So everyone you know Amazon web

Okay. So everyone you know Amazon web services we have we are a subsidiary of

services we have we are a subsidiary of amazon.com and for amazon.com uh we have

amazon.com and for amazon.com uh we have a global like inventory logistics

a global like inventory logistics services providers so you can see we

services providers so you can see we heavily use a lot of IoT services

heavily use a lot of IoT services already within our Amazon uh business

already within our Amazon uh business for example in our logistic fulfillment

for example in our logistic fulfillment centers there are 520k plus of robotic

centers there are 520k plus of robotic drive units

drive units And these drive units already supporting

And these drive units already supporting three over 300 facilities. So they run

three over 300 facilities. So they run autonomously and they run with a lots of

autonomously and they run with a lots of sensors and also like pre configured

sensors and also like pre configured like RPA or routes etc. And second one,

like RPA or routes etc. And second one, we have a large fleets of drivers and

we have a large fleets of drivers and carriers and there's a lot of

carriers and there's a lot of interaction between how we keep track on

interaction between how we keep track on the on the route, how we keep track on

the on the route, how we keep track on the packages etc. And then further we

the packages etc. And then further we have we also have Amazon go store which

have we also have Amazon go store which is a physical store and this physical

is a physical store and this physical store we emphasize on our latest

store we emphasize on our latest technology uh just walk out

technology uh just walk out technologies. So it's simply keep track

technologies. So it's simply keep track on how you go into the shop and then

on how you go into the shop and then with a QR code you scan it and then you

with a QR code you scan it and then you can directly go into bringing out

can directly go into bringing out whatever ro uh groceries you have and

whatever ro uh groceries you have and then you just walk out and then the uh

then you just walk out and then the uh our technologies will help to validate

our technologies will help to validate what kind of groceries you have uh bring

what kind of groceries you have uh bring it out and then we automatically check

it out and then we automatically check out for you. So these type of

out for you. So these type of innovations actually we leverage heavily

innovations actually we leverage heavily on our IoT uh services and also uh like

on our IoT uh services and also uh like the concept the ideas and we make it

the concept the ideas and we make it into realities. So this is only the part

into realities. So this is only the part of it uh how we leverage IoT innovations

of it uh how we leverage IoT innovations for Amazon itself

for Amazon itself especially for our customers. We do have

especially for our customers. We do have a lot of global customers uh focusing on

a lot of global customers uh focusing on different sectors. They also leverage

different sectors. They also leverage IoT uh into their business like for

IoT uh into their business like for example in like the consumer units uh

example in like the consumer units uh like wise actually it they quickly build

like wise actually it they quickly build and secure products unlock data

and secure products unlock data connectivity there a lot of things they

connectivity there a lot of things they are ongoing to do in order to streamline

are ongoing to do in order to streamline the whole development process such that

the whole development process such that they can lower the cost of producing

they can lower the cost of producing each of the units of their cameras. Also

each of the units of their cameras. Also in terms of like industrial

in terms of like industrial manufacturing there are a lot of ways to

manufacturing there are a lot of ways to monitor each type of the process and

monitor each type of the process and each of the uh each of the the system

each of the uh each of the the system and also the factory units uh etc. Uh in

and also the factory units uh etc. Uh in terms of automobiles we have customers

terms of automobiles we have customers they focus on connected vehicles. So the

they focus on connected vehicles. So the vehicles uh they comes with sensor they

vehicles uh they comes with sensor they comes with network and then it can

comes with network and then it can monitor in real time how does it go and

monitor in real time how does it go and also especially for EW uh you can also

also especially for EW uh you can also monitor about the like the batteries you

monitor about the like the batteries you can monitor about the mileage etc

can monitor about the mileage etc there's a lot of things they can keep

there's a lot of things they can keep connected to and all these are powered

connected to and all these are powered by the IoT services

by the IoT services so what I what I want to emphasize with

so what I what I want to emphasize with these of the use cases is that uh to put

these of the use cases is that uh to put it into your scenario on like thinking

it into your scenario on like thinking about which type of uh services or which

about which type of uh services or which type of products you want to focus on.

type of products you want to focus on. The first thing is you need to identify

The first thing is you need to identify the use case very very clearly like what

the use case very very clearly like what are the use case what are the market or

are the use case what are the market or what are the industry you are going to

what are the industry you are going to target to and then in that industry or

target to and then in that industry or in that market what are the target

in that market what are the target audience what are the target customers

audience what are the target customers or users they have they are and then you

or users they have they are and then you are going to work backward and then

are going to work backward and then craft the personas and then to crafting

craft the personas and then to crafting the uh the details of the use cases how

the uh the details of the use cases how can it be used and in what sense uh the

can it be used and in what sense uh the IoT can take plays in like for example

IoT can take plays in like for example which type of sensors uh are needed and

which type of sensors uh are needed and then you work backward with okay if I

then you work backward with okay if I use this like for example smart building

use this like for example smart building there's a lot of sensors like measuring

there's a lot of sensors like measuring the temperature measuring the humidity

the temperature measuring the humidity measuring a lot of things so what which

measuring a lot of things so what which type of the sensor or which type of the

type of the sensor or which type of the matrix actually you need to monitor and

matrix actually you need to monitor and then how you grabs the insight and this

then how you grabs the insight and this way you need to think about the hypo

way you need to think about the hypo hypothesis and everything we talk about

hypothesis and everything we talk about hypothesis you need to make experiment

hypothesis you need to make experiment in order to test it and then finally to

in order to test it and then finally to whether they whether your idea is

whether they whether your idea is bulletproof or not. So in that sense it

bulletproof or not. So in that sense it is uh traditionally this process is

is uh traditionally this process is really a long journey to go especially

really a long journey to go especially when you go out and then you try to go

when you go out and then you try to go into a building for example and then you

into a building for example and then you try to identify the use cases and then

try to identify the use cases and then you try to grab okay whether this use

you try to grab okay whether this use cases is a good problem to solve or not

cases is a good problem to solve or not and then you are gradually to think okay

and then you are gradually to think okay this is this comes with a values no

this is this comes with a values no matter it is a business value no no

matter it is a business value no no matter This is um like uh like like what

matter This is um like uh like like what Edward has has said before is about the

Edward has has said before is about the SG objectives or the values uh social

SG objectives or the values uh social values etc. So it does have to have a

values etc. So it does have to have a values in order to prove your problem is

values in order to prove your problem is worth to solve. After you identify all

worth to solve. After you identify all these you need to go through experiment

these you need to go through experiment and then traditionally you go through

and then traditionally you go through experiment you need to write a lot of

experiment you need to write a lot of codes you need to streamline all the

codes you need to streamline all the physical components hardware and

physical components hardware and softwares etc. But there are a lot of

softwares etc. But there are a lot of development work to do. So this only the

development work to do. So this only the experiment. This is not the production

experiment. This is not the production grade and this is only this already

grade and this is only this already takes up a lot of time for you to

takes up a lot of time for you to validate your ideas it is good or not.

validate your ideas it is good or not. But in this AI era uh we talk a lot of

But in this AI era uh we talk a lot of how AI can help you to work on this

how AI can help you to work on this stuff. That's why we want you to

stuff. That's why we want you to introduce and also accelerates the

introduce and also accelerates the process of experimentation with agentic

process of experimentation with agentic AI. So what specifically uh on agentic

AI. So what specifically uh on agentic AI can helps in this part of it? For

AI can helps in this part of it? For example, uh maybe you work work along

example, uh maybe you work work along with a lot of chatboards or a lot of LOM

with a lot of chatboards or a lot of LOM actually you uh prom with a questions

actually you uh prom with a questions and then ask okay what are the ideas

and then ask okay what are the ideas what are the details uh can you come up

what are the details uh can you come up with a uh comprehensive ideas this is

with a uh comprehensive ideas this is only the start of how you're going to

only the start of how you're going to start to validate your ideas and start

start to validate your ideas and start to do the experimentation but a genic AI

to do the experimentation but a genic AI can do even more than that

can do even more than that so we talk about like AI is changing

so we talk about like AI is changing software especially it's changing how

software especially it's changing how software being made so um a little bit

software being made so um a little bit history about AI like back to 20 2024

history about AI like back to 20 2024 actually uh we only comes with assistant

actually uh we only comes with assistant so the assistant can help developers to

so the assistant can help developers to write code faster like for example it um

write code faster like for example it um you have the features of autocomplete it

you have the features of autocomplete it try to predict what are your next

try to predict what are your next command or next uh code snippets you

command or next uh code snippets you want to generate and that's it. But

want to generate and that's it. But gradually with the evolving of AI

gradually with the evolving of AI technologies in 2025, it already comes

technologies in 2025, it already comes with agents. So we talk a lot of things

with agents. So we talk a lot of things a of agents I think in the mid of 2025

a of agents I think in the mid of 2025 and then it gradually evolved into a

and then it gradually evolved into a more smarter a smarter more intelligence

more smarter a smarter more intelligence agents that can complete the task end to

agents that can complete the task end to end but it's still with human in the

end but it's still with human in the loop and we predict that in 2026 which

loop and we predict that in 2026 which is this year it become more autonomous

is this year it become more autonomous like the agents itself can have a smart

like the agents itself can have a smart or intell intelligence in order to

or intell intelligence in order to complete the development task end to end

complete the development task end to end and with bounded independency.

and with bounded independency. So what what are the boundary

So what what are the boundary independence like you may or may not get

independence like you may or may not get involved into the development task or in

involved into the development task or in involved in the task it's up to you okay

involved in the task it's up to you okay but usually we talk about agentic

but usually we talk about agentic developments there is a lot of thoughts

developments there is a lot of thoughts uh I want to share case like on two

uh I want to share case like on two flips on agentic developments so on one

flips on agentic developments so on one flips on pros actually is it's great

flips on pros actually is it's great because it's autonomies uh like agency

because it's autonomies uh like agency take and complete increasingly

take and complete increasingly challenging tasks autonomously. And also

challenging tasks autonomously. And also uh it can helps with true collaborations

uh it can helps with true collaborations like developers and agent right now is

like developers and agent right now is work together and get more to done and

work together and get more to done and also higher quality because it can

also higher quality because it can benchmark with a lot of references, a

benchmark with a lot of references, a lot of materials, a lot of knowledge

lot of materials, a lot of knowledge available on the internet actually they

available on the internet actually they can come like supporting you with a

can come like supporting you with a higher quality products. But on the flip

higher quality products. But on the flip side uh aenic AI accelerate all these

side uh aenic AI accelerate all these stuff but it also back into a problem

stuff but it also back into a problem that you need to maintain the quality.

that you need to maintain the quality. You need to have a certain control over

You need to have a certain control over uh your tools how to being used it or

uh your tools how to being used it or your code uh rep repository qualities

your code uh rep repository qualities and also how this can be scale instead

and also how this can be scale instead of just I use it in a personal uh

of just I use it in a personal uh computers I just talk onetoone but

computers I just talk onetoone but without any collaboration without uh

without any collaboration without uh with your colleagues. So all these kind

with your colleagues. So all these kind of things actually we

of things actually we acknowledge that is a two flips and we

acknowledge that is a two flips and we try to solve it and we try to improve it

try to solve it and we try to improve it such that uh your productivity to

such that uh your productivity to accelerate experimentations is more

accelerate experimentations is more smooth or smoother.

smooth or smoother. So here comes that question is how would

So here comes that question is how would the development experience look like if

the development experience look like if we could take full advantage of working

we could take full advantage of working with AI agents.

with AI agents. This is a visions and this is how we

This is a visions and this is how we comes up with Kira IDE such that we can

comes up with Kira IDE such that we can help uh and empower a lot of developers

help uh and empower a lot of developers globally to work uh seamlessly with AI

globally to work uh seamlessly with AI agents.

agents. So that's why we have a genetic id and

So that's why we have a genetic id and this is our logos of of a ghost that's

this is our logos of of a ghost that's cur

cur uh I want to echo what our VP and CTO of

uh I want to echo what our VP and CTO of Amazon.com has stated during our last

Amazon.com has stated during our last year reinvent which is uh which just uh

year reinvent which is uh which just uh hold on last November on 2025

hold on last November on 2025 uh we emphasize that there are only all

uh we emphasize that there are only all new inventions and you need experiment

new inventions and you need experiment and you need to be willing to fail fail

and you need to be willing to fail fail and be gently corrected and all this

and be gently corrected and all this process actually you can accelerate with

process actually you can accelerate with kuro and also other agentic IDE or AI.

kuro and also other agentic IDE or AI. So here you may see this uh screen or

So here you may see this uh screen or you may come across with the name of

you may come across with the name of kuro IDE uh if you have experience on

kuro IDE uh if you have experience on coding programming and actually this is

coding programming and actually this is a very common uh interfaces that you

a very common uh interfaces that you come across. So what specifically

come across. So what specifically features that can help uh on

features that can help uh on accelerating your experimentations

accelerating your experimentations and first of all I would like to

and first of all I would like to introduce a uh features about agent

introduce a uh features about agent hooks. So in terms of agent hooks uh

hooks. So in terms of agent hooks uh actually you can set some automatically

actually you can set some automatically autonomously conditions that uh you when

autonomously conditions that uh you when you interact with the ID like for

you interact with the ID like for example you write a new code or you ask

example you write a new code or you ask the agents to create a new files or new

the agents to create a new files or new uh code snippets actually you can

uh code snippets actually you can trigger uh the hooks that you

trigger uh the hooks that you preconfigured here for example if I ask

preconfigured here for example if I ask agents to create a new features and pro

agents to create a new features and pro possibly they it generate a several

possibly they it generate a several coding files or the empty files and then

coding files or the empty files and then I can preconfig a hook that uh they need

I can preconfig a hook that uh they need to keep track on the change log like for

to keep track on the change log like for example what are the files that being

example what are the files that being writ

in in such a way actually it can help to minimize uh many work or many overhead

minimize uh many work or many overhead had uh advers that you need to be done

had uh advers that you need to be done uh previously and it is more than that.

uh previously and it is more than that. Uh it's up to you on how you're going to

Uh it's up to you on how you're going to chain uh the actions or interactions to

chain uh the actions or interactions to streamline your own process and this is

streamline your own process and this is one of the features that supported in

one of the features that supported in Kuro ID.

Kuro ID. Okay, for the second one is about we

Okay, for the second one is about we improve on the context management. If

improve on the context management. If you have experiences in using uh any

you have experiences in using uh any kind of IDE or uh like a gentic IDE uh

kind of IDE or uh like a gentic IDE uh when you talk with the agents or talk

when you talk with the agents or talk with uh the LOM uh within the IDE

with uh the LOM uh within the IDE actually you may come across with some

actually you may come across with some problems about the context window. uh

problems about the context window. uh maybe you generate a sequence of files

maybe you generate a sequence of files will you become increasingly large of

will you become increasingly large of the of the code repositories and then

the of the code repositories and then you will come across with a error like

you will come across with a error like for example uh it reach reaches the

for example uh it reach reaches the maximum context window. So what kind of

maximum context window. So what kind of context window is this? Context window

context window is this? Context window you can imagine it is a memory uh that

you can imagine it is a memory uh that the agentic or the agents can take place

the agentic or the agents can take place such that they can cross reference with

such that they can cross reference with uh the work you've done before or the

uh the work you've done before or the chat history you done before but we

chat history you done before but we built in this context management into ko

built in this context management into ko ID such that you don't need to preconfig

ID such that you don't need to preconfig all the like memory stuffs or other

all the like memory stuffs or other stuffs uh when it reaches the maximum

stuffs uh when it reaches the maximum context window actually you can uh we

context window actually you can uh we will handle it for you. When you reach

will handle it for you. When you reach the maximum context window actually we

the maximum context window actually we will summarize uh the your conversation

will summarize uh the your conversation before and what are the interactions you

before and what are the interactions you uh work before and then it will cascade

uh work before and then it will cascade that kind of context into a new chat

that kind of context into a new chat such that the agent can continuous with

such that the agent can continuous with what you wor with him or work with the

what you wor with him or work with the agents before. such that you will

agents before. such that you will experience a very streamlined uh user

experience a very streamlined uh user experiences.

Okay. So the next one is about timeline checkpointing. So when you go and work

checkpointing. So when you go and work with the uh agents uh subsequently

with the uh agents uh subsequently there's a lot of things that ongoing

there's a lot of things that ongoing like for example you are going to create

like for example you are going to create several files or you are going to create

several files or you are going to create a lot of codes during your interactions

a lot of codes during your interactions and we automatically add checkpoint on

and we automatically add checkpoint on each of the interactions you made with

each of the interactions you made with kirao. So why this is so important?

kirao. So why this is so important? Because for some cases the agent might

Because for some cases the agent might not generate the things you want to do

not generate the things you want to do or the the agent may go wrong. So the

or the the agent may go wrong. So the checkpoint here can help you to easily

checkpoint here can help you to easily reverse and then go back to uh the work

reverse and then go back to uh the work you've done before and then you can

you've done before and then you can start there fine-tune your prompt to the

start there fine-tune your prompt to the agents and continuous working with the

agents and continuous working with the time checkpoint you want to start with.

time checkpoint you want to start with. Uh this is a very important features

Uh this is a very important features that we come up with our customers

that we come up with our customers requests. uh because when you go with

requests. uh because when you go with other like agentic ID or agents that

other like agentic ID or agents that help you to develop quotes actually one

help you to develop quotes actually one of the very important or like um problem

of the very important or like um problem is that the people they they are

is that the people they they are difficult to control and manage how the

difficult to control and manage how the agent works and that's why we come up

agent works and that's why we come up with these features and editing into Kro

with these features and editing into Kro ID in order to help you to develop the

ID in order to help you to develop the products smoothly.

Okay. And finally which is very important

And finally which is very important concept is on the whole curo IDE we

concept is on the whole curo IDE we focus on two different mode. One is

focus on two different mode. One is white coding and the other one is

white coding and the other one is specdriven development. So everyone you

specdriven development. So everyone you may come across with white coding is

may come across with white coding is that you simply put in a natural uh or

that you simply put in a natural uh or layman terms uh into an agents and then

layman terms uh into an agents and then the agent will start to develop a lot of

the agent will start to develop a lot of stuff to it for you and you have little

stuff to it for you and you have little control on tracing what this does but

control on tracing what this does but spec driven development uh is more

spec driven development uh is more important in order to ship the products

important in order to ship the products that go align with the SDLC practices.

that go align with the SDLC practices. So you you can see the uh very simple

So you you can see the uh very simple flowchart here. It start from planning

flowchart here. It start from planning design and then you will go through

design and then you will go through implementations and then you will go to

implementations and then you will go to testing and QA and then if there

testing and QA and then if there anything is to find actually it will

anything is to find actually it will reverse back to the implementations and

reverse back to the implementations and then to refine or debug uh what are the

then to refine or debug uh what are the product what are the features is and

product what are the features is and then you're going forward with the

then you're going forward with the fine-tuning or refined version of the of

fine-tuning or refined version of the of the codes and then you go to testing and

the codes and then you go to testing and QA and then they approved it and you're

QA and then they approved it and you're going to deployment and And finally the

going to deployment and And finally the product is being launched and with

product is being launched and with maintenance. So this is a very typical

maintenance. So this is a very typical traditional SDLC uh software development

traditional SDLC uh software development life cycle practices and you can see we

life cycle practices and you can see we identify why coding is only a part of it

identify why coding is only a part of it uh during the implementations but for

uh during the implementations but for specdriven development we cover the

specdriven development we cover the whole traditional SDLC practices from

whole traditional SDLC practices from planning design implementation testing

planning design implementation testing QA and deployment. So which this

QA and deployment. So which this approach help a lot of our customers to

approach help a lot of our customers to implement and also accelerate the whole

implement and also accelerate the whole process with ko.

process with ko. So here you can see this is uh our main

So here you can see this is uh our main screen on how to choose whether you are

screen on how to choose whether you are going for w coding or going to

going for w coding or going to specdriven development and you can see a

specdriven development and you can see a lot of things. uh if you're going to

lot of things. uh if you're going to choose a specriven development there's a

choose a specriven development there's a three key documents being generated the

three key documents being generated the one is about requirements the

one is about requirements the requirements actually is a business

requirements actually is a business requirements so there's a very layman

requirements so there's a very layman terms that going to expand uh your ass

terms that going to expand uh your ass on like for example you de you want to

on like for example you de you want to develop a products uh you want to

develop a products uh you want to develop like for example a e-commerce

develop like for example a e-commerce portal and then they were going to

portal and then they were going to expand this with a business requirement

expand this with a business requirement documentations and then you are going to

documentations and then you are going to pro going for a subsequent process uh in

pro going for a subsequent process uh in order to review that business

order to review that business requirement. Is that truly reflects what

requirement. Is that truly reflects what you want to test it or try it and then

you want to test it or try it and then you move forward to design phase uh that

you move forward to design phase uh that generate and summarize uh how the high

generate and summarize uh how the high level design it is and once you got that

level design it is and once you got that design documents is ready and then you

design documents is ready and then you can go beyond that to generate a task

can go beyond that to generate a task list which is considered above the

list which is considered above the actual implementation task. So here this

actual implementation task. So here this approach can help the teams and also you

approach can help the teams and also you to keep trace and keep track on every

to keep trace and keep track on every part of it in order to keep trace like

part of it in order to keep trace like from the requirements to the design and

from the requirements to the design and then to the implementation task.

then to the implementation task. So this is one of it

and we talk about prototype to production if you have aware of what I

production if you have aware of what I mentioned about white coding and also

mentioned about white coding and also spectriff development. So you can easily

spectriff development. So you can easily see the differences between that and

see the differences between that and here I just want to identify some part

here I just want to identify some part of it like for prototyping actually

of it like for prototyping actually there are some of the characteristics

there are some of the characteristics like for example it's a fast delivery

like for example it's a fast delivery and also usually it only involve a unit

and also usually it only involve a unit approach that can enable you to quickly

approach that can enable you to quickly validate your ideas and usually for

validate your ideas and usually for prototype is is focusing on a single

prototype is is focusing on a single value propositions uh instead of a

value propositions uh instead of a multiple value proposition or a broad a

multiple value proposition or a broad a general product value propositions

general product value propositions and in terms of time frame it is it

and in terms of time frame it is it should be achieved in a near-time

should be achieved in a near-time objective. It is not a months or weeks

objective. It is not a months or weeks or years objective. It should be in an

or years objective. It should be in an hours or days objectives and usually

hours or days objectives and usually what team size involved it is about like

what team size involved it is about like for example it's only a single person

for example it's only a single person you want to validate whether this is uh

you want to validate whether this is uh achievable or not or even a very small

achievable or not or even a very small team size uh like for example in in AWS

team size uh like for example in in AWS actually we we focusing on two visa team

actually we we focusing on two visa team which uh we identify the small team uh

which uh we identify the small team uh structure that is the most efficient way

structure that is the most efficient way to work on a project and vice versa For

to work on a project and vice versa For production grade products, uh it usually

production grade products, uh it usually involve a long-term planning uh very

involve a long-term planning uh very detailed planning, detailed

detailed planning, detailed documentations, what it does, what uh

documentations, what it does, what uh what are the design documents is and

what are the design documents is and also what are the actual tasks you need

also what are the actual tasks you need to be done in a sequential order, what

to be done in a sequential order, what are the dependency for each of of it.

are the dependency for each of of it. That's why it comes with a very

That's why it comes with a very comprehensive feature sets and these

comprehensive feature sets and these comprehensive feature sets also

comprehensive feature sets also corresponding to the value propositions

corresponding to the value propositions uh that delivered throughout the

uh that delivered throughout the products and usually for the objective

products and usually for the objective it closely related and correlated to the

it closely related and correlated to the actual business objective and usually it

actual business objective and usually it involve a larger team size. So these are

involve a larger team size. So these are the characteristic between prototype and

the characteristic between prototype and productions

productions and what are what are the differences

and what are what are the differences between that. Uh for prototype we

between that. Uh for prototype we strongly suggest you to leverage uh ko

strongly suggest you to leverage uh ko uh in terms of the whiteboard

uh in terms of the whiteboard [clears throat] it can quickly validate

[clears throat] it can quickly validate your ideas. So it's naming terms you

your ideas. So it's naming terms you just type it in uh and then you will try

just type it in uh and then you will try to generate a very simple uh structure

to generate a very simple uh structure of code repository and then uh it will

of code repository and then uh it will comes with a quick start and then you

comes with a quick start and then you can quickly uh deploying and also test

can quickly uh deploying and also test it with the local machines and then you

it with the local machines and then you can quickly get back the feedbacks

can quickly get back the feedbacks whether this is work or not whether this

whether this is work or not whether this is the experience you wanted or not etc.

is the experience you wanted or not etc. it would directly generate the cooks for

it would directly generate the cooks for a task instead of doing a long-term

a task instead of doing a long-term planning or or detailed business

planning or or detailed business requirement etc.

requirement etc. Vice versa, if this is a specd driven

Vice versa, if this is a specd driven mode, actually you can see there's a lot

mode, actually you can see there's a lot of detailed tasks being created and then

of detailed tasks being created and then it will go through requirements design

it will go through requirements design and task documents subsequentially

and task documents subsequentially and you can see the agent will follow

and you can see the agent will follow each of the tasks uh being numbers. You

each of the tasks uh being numbers. You can see there's a lot of numbers 1 2 3 4

can see there's a lot of numbers 1 2 3 4 5 and then you would break down into a

5 and then you would break down into a subtask with 1.1 1.2 and 1.3 etc. So it

subtask with 1.1 1.2 and 1.3 etc. So it is very very clearly defined on the ask

is very very clearly defined on the ask with the specs property. So here I want

with the specs property. So here I want to I want to also emphasize uh there's

to I want to also emphasize uh there's some of the concept about specs. So

some of the concept about specs. So everyone you when you hear about spec

everyone you when you hear about spec specifications maybe you can come across

specifications maybe you can come across with uh some of the detailed terms like

with uh some of the detailed terms like uh what are what what what are the

uh what are what what what are the actual matrix you want to measure or

actual matrix you want to measure or what are the actual features you want to

what are the actual features you want to create what are the expected behavior of

create what are the expected behavior of that features etc. And that exactly we

that features etc. And that exactly we build it in into the specdriven mode is

build it in into the specdriven mode is that we define clearly on the specs

that we define clearly on the specs property what uh what are needed to be

property what uh what are needed to be achieved it. Like for example instead of

achieved it. Like for example instead of we saying uh just generate an interface

we saying uh just generate an interface for me uh to interact with uh with specs

for me uh to interact with uh with specs property uh we will staying that

property uh we will staying that uh like for example when a user comes in

uh like for example when a user comes in to the homepage

to the homepage then uh the interface should comes up

then uh the interface should comes up within for example 0.5 seconds and then

within for example 0.5 seconds and then to render their latest uh dashboard

to render their latest uh dashboard units for example

units for example So these are the actual specification

So these are the actual specification with specs property and then we build it

with specs property and then we build it in into the agent in Kirro IDE such that

in into the agent in Kirro IDE such that when the agent are generating the

when the agent are generating the business requirement documentations or

business requirement documentations or the design documentation or even the

the design documentation or even the task uh task list it will have the

task uh task list it will have the knowledge and the mindset to follow

knowledge and the mindset to follow these kind of approach in order to

these kind of approach in order to generate a more clear specifications

generate a more clear specifications with property.

with property. And then for the task list like what I

And then for the task list like what I mentioned before we have a layered task

mentioned before we have a layered task list there's a general task or even

list there's a general task or even there's by phases and then we'll break

there's by phases and then we'll break down into a subtask in order for the

down into a subtask in order for the agents to go through it one by one and

agents to go through it one by one and then those tasks with uh sequence

then those tasks with uh sequence actually we consider about the

actually we consider about the dependency already such that they will

dependency already such that they will build from the fundamentals and then

build from the fundamentals and then they will build uh one layer up and then

they will build uh one layer up and then on there etc. So this is already the

on there etc. So this is already the step-by-step approach for the

step-by-step approach for the implementation.

So here is just a highlight for you uh what are the requirements documentation

what are the requirements documentation is and then what are the design

is and then what are the design documentation is and what are the task

documentation is and what are the task list documentation is.

list documentation is. I would like to jump into a quick demo

I would like to jump into a quick demo on uh what I just achieved uh from

on uh what I just achieved uh from yesterday.

yesterday. So we go up to this. Ah let me show you

So we go up to this. Ah let me show you the actual product first.

the actual product first. Okay. So here you can see uh we have a

Okay. So here you can see uh we have a very simple straightforward web portal.

very simple straightforward web portal. Uh I just picked two common utility

Uh I just picked two common utility service proto uh to generate like for

service proto uh to generate like for example I want to convert an image uh

example I want to convert an image uh from one kind of supporting format to

from one kind of supporting format to another supporting format. It's very

another supporting format. It's very very easy and uh common utilities you

very easy and uh common utilities you you want to leverage it and I want to

you want to leverage it and I want to test it whether they whether the

test it whether they whether the experience it is. So I just like for

experience it is. So I just like for example I choose file and uh like for

example I choose file and uh like for example I choose this one is a kira logo

example I choose this one is a kira logo and then it comes with preview which is

and then it comes with preview which is nice and then I want to output it into a

nice and then I want to output it into a like for example PDF file and I just

like for example PDF file and I just want don't want other conversion options

want don't want other conversion options and then I just simply click convert

and then I just simply click convert images and it's done and then when I

images and it's done and then when I download convert images I open it it

download convert images I open it it successfully go through the PDF file

successfully go through the PDF file here. Yeah. Oh, it is white in

here. Yeah. Oh, it is white in background but never mind. I already

background but never mind. I already converted into beta file as simple as

converted into beta file as simple as that. So also I also uh quickly create a

that. So also I also uh quickly create a QR code generators like using the

QR code generators like using the leverage um the libraries and you'll see

leverage um the libraries and you'll see there's a nice QR code generator with

there's a nice QR code generator with self-bait logos and here is just a front

self-bait logos and here is just a front end portal and then at the back end I

end portal and then at the back end I also leverage an open API standard in

also leverage an open API standard in order to create the documentations such

order to create the documentations such that I can quickly test it and whether

that I can quickly test it and whether they whether uh the API endpoints are

they whether uh the API endpoints are work or not. So you can see at the back

work or not. So you can see at the back end we have several image conversion um

end we have several image conversion um uh image conversion APIs and then I can

uh image conversion APIs and then I can directly test it here and then etc. So

directly test it here and then etc. So very very simple uh portals

very very simple uh portals experiences and I do it with Kira and

experiences and I do it with Kira and then I just start all these stuff like

then I just start all these stuff like yesterday when I just finished all the

yesterday when I just finished all the meetings and then I just uh tell Kira

meetings and then I just uh tell Kira about okay you quickly design the

about okay you quickly design the portals uh with uh image conversion

portals uh with uh image conversion first and then you can see here we have

first and then you can see here we have a ko specifications folders and then it

a ko specifications folders and then it contain uh different asks for me and

contain uh different asks for me and then with different RS actually is

then with different RS actually is layered down to a requirements

layered down to a requirements documentations uh the design

documentations uh the design documentations and also task

documentations and also task documentations.

documentations. Okay. So when we look into each of the

Okay. So when we look into each of the part of it uh the requirement

part of it uh the requirement documentation is very very naming terms

documentation is very very naming terms just outlining what are the business

just outlining what are the business requirements uh uh that going going to

requirements uh uh that going going to be achieved and then you will break down

be achieved and then you will break down into several requirement and also

into several requirement and also there's acceptance criteria and when you

there's acceptance criteria and when you look closely into acceptance criteria

look closely into acceptance criteria there is the specs property I just

there is the specs property I just mentioned for example the acceptance

mentioned for example the acceptance criteria here is when a client sends a

criteria here is when a client sends a post request with a valid image file in

post request with a valid image file in a supporting im input format the API

a supporting im input format the API server shall accept the upload and

server shall accept the upload and return SSS respond. So it's clearly

return SSS respond. So it's clearly defined what are what are the situations

defined what are what are the situations it's and also what are the acceptance

it's and also what are the acceptance criteria or the intended behavior from

criteria or the intended behavior from that features.

that features. So here you can see all these

So here you can see all these requirement documentations follow the

requirement documentations follow the same approach and uh for business

same approach and uh for business generating business requirements you can

generating business requirements you can first generate it and then you can try

first generate it and then you can try to review it with uh like or manually or

to review it with uh like or manually or you through the agent to review on your

you through the agent to review on your own and then when you feel satisfied

own and then when you feel satisfied with that requirements documentations

with that requirements documentations and they will go through it to the

and they will go through it to the second stage about the design file and

second stage about the design file and the design file is about a high level

the design file is about a high level technical

technical And you can see we got into a one more

And you can see we got into a one more layer about the technologies for example

layer about the technologies for example architectures about the technology set

architectures about the technology set being used and then some of the uh

being used and then some of the uh architecture diagrams or even uh the

architecture diagrams or even uh the APIs structures etc. So you generate the

APIs structures etc. So you generate the high level design with like for example

high level design with like for example data type etc a lot of things about high

data type etc a lot of things about high level technical design and follow the

level technical design and follow the same approach like what requirements

same approach like what requirements does actually you can reveal the whole

does actually you can reveal the whole design files and then you feel satisfied

design files and then you feel satisfied and then you can try to proceed with a

and then you can try to proceed with a task list and here is a actual

task list and here is a actual implementation plan task list and then

implementation plan task list and then being numbered with a sequence. You can

being numbered with a sequence. You can see it uh it usually start with a setup

see it uh it usually start with a setup project structure with a configurations

project structure with a configurations and usually it is a initi initiation or

and usually it is a initi initiation or the uh like package or etc. And the most

the uh like package or etc. And the most importantly is you can easily keep trace

importantly is you can easily keep trace on what are the tasks already being

on what are the tasks already being done. And if the task being completed

done. And if the task being completed you can directly see in the task list

you can directly see in the task list that the task being completed and if

that the task being completed and if with each of the task actually it will

with each of the task actually it will corresponding to the business

corresponding to the business requirements you specify or you revealed

requirements you specify or you revealed before like for example for the task two

before like for example for the task two actually it corresponding to uh 8.1 8.2

actually it corresponding to uh 8.1 8.2 2 8.3 8.5 uh actual business

2 8.3 8.5 uh actual business requirements that specified in the

requirements that specified in the earlier requirements documentations.

earlier requirements documentations. So this kind of approach we call it as

So this kind of approach we call it as spec driven development approach and it

spec driven development approach and it can help your team or even you in person

can help your team or even you in person to keep track on and keep trace on the

to keep track on and keep trace on the overall uh uh project implementations

overall uh uh project implementations easily.

easily. Okay.

Okay. So

So I would strongly suggest you to um like

I would strongly suggest you to um like try to

try to depends on your use cases identify uh we

depends on your use cases identify uh we have we comes with two different modes.

have we comes with two different modes. The first mode is about the white coding

The first mode is about the white coding and then the white coding is good for

and then the white coding is good for you to try to validate your ideas. You

you to try to validate your ideas. You try to start building something quick in

try to start building something quick in order uh something small in order to

order uh something small in order to validate whether your ideas work or not.

validate whether your ideas work or not. put it into a IoT uh use cases uh like

put it into a IoT uh use cases uh like for example you usually handle uh

for example you usually handle uh several protocols from uh the sensors

several protocols from uh the sensors from the hardwares etc and also you may

from the hardwares etc and also you may come across with the requirements you

come across with the requirements you want to easily visualize what it does or

want to easily visualize what it does or what are the sensor insights etc. So you

what are the sensor insights etc. So you can leverage Kirro IDE in order to test

can leverage Kirro IDE in order to test it, validate it, whether uh the insights

it, validate it, whether uh the insights you generate from your hypothesis is

you generate from your hypothesis is correct or not or even you can ask the

correct or not or even you can ask the intelligence from the agents about okay

intelligence from the agents about okay what kind of insight I can generate from

what kind of insight I can generate from those sensor or those matrix. uh it will

those sensor or those matrix. uh it will comes up with some of some of the

comes up with some of some of the suggestions and then you try to mix and

suggestions and then you try to mix and match those suggestion in order to see

match those suggestion in order to see whether this comes up with a business

whether this comes up with a business where that uh come that go align with

where that uh come that go align with your use cases or in the target use uh

your use cases or in the target use uh target use cases or target customers and

target use cases or target customers and whatever you you validated that idea is

whatever you you validated that idea is a good to go and then you can try to use

a good to go and then you can try to use the specdriven development in order to

the specdriven development in order to comprehensively build your products into

comprehensively build your products into a

a uh concrete structures with identified

uh concrete structures with identified business requirements with a high level

business requirements with a high level technical design documents and all those

technical design documents and all those the task list that can guide the agents

the task list that can guide the agents can help you to build alongside with

can help you to build alongside with you. So this will be the approach uh I

you. So this will be the approach uh I would strongly suggest you to go for uh

would strongly suggest you to go for uh from prototype to production and

from prototype to production and especially some tricks uh if you go for

especially some tricks uh if you go for production

production um you can advise the agents to follow

um you can advise the agents to follow some of the standards uh that available

some of the standards uh that available um you can easily search on the internet

um you can easily search on the internet what how a production products or

what how a production products or technical software products can go live

technical software products can go live with and usually this comes with several

with and usually this comes with several major components. like for example a

major components. like for example a very structural software archite

very structural software archite architectures or even in a microservices

architectures or even in a microservices approach and then you can go a live with

approach and then you can go a live with some of the standard like for example if

some of the standard like for example if you're going for APIs uh you may ask the

you're going for APIs uh you may ask the agents to follow the open API uh

agents to follow the open API uh standard or you follow uh the swagger uh

standard or you follow uh the swagger uh API standards and these kind of things

API standards and these kind of things can help you to make your products

can help you to make your products easily being traced it and then you can

easily being traced it and then you can also put a lot of other features just

also put a lot of other features just tell the uh agents to do on your own for

tell the uh agents to do on your own for example put it uh monitoring matrix

example put it uh monitoring matrix observabilities

observabilities uh even API key management etc there's a

uh even API key management etc there's a lot of common services you can you can

lot of common services you can you can add it in and uh traditionally you if

add it in and uh traditionally you if you need to add these type of modules it

you need to add these type of modules it would takes like days weeks in order to

would takes like days weeks in order to fine-tune your code and refine it for

fine-tune your code and refine it for you but with agents actually you can

you but with agents actually you can done like like hours or days. So you can

done like like hours or days. So you can easily chip in into it and then try to

easily chip in into it and then try to see how the uh how the effect it is.

see how the uh how the effect it is. Okay. So after all um uh it is just a

Okay. So after all um uh it is just a quick demo on uh how Kero ID can work uh

quick demo on uh how Kero ID can work uh on both prototypes or the production

on both prototypes or the production grade products and I would strongly

grade products and I would strongly suggest you register for KO today uh via

suggest you register for KO today uh via that link or even the QR code I just

that link or even the QR code I just generated with our with my common

generated with our with my common services uh portals. Uh feel free to go

services uh portals. Uh feel free to go directly onto it. uh when you are the

directly onto it. uh when you are the first time to use Kro actually we comes

first time to use Kro actually we comes with uh extra and bonus credits for you

with uh extra and bonus credits for you to run your project on it. So just give

to run your project on it. So just give a try on this.

a try on this. Okay. And secondly we just a little

Okay. And secondly we just a little promotion on uh our event. So we do have

promotion on uh our event. So we do have a lot of new features coming in uh from

a lot of new features coming in uh from our last AWS reinvent and we have a

our last AWS reinvent and we have a online events uh that target to both

online events uh that target to both technical and business professionals

technical and business professionals that want to seek and exploring uh our

that want to seek and exploring uh our latest innovations. Especially we uh on

latest innovations. Especially we uh on Loom we announced there's Amazon Nova 2

Loom we announced there's Amazon Nova 2 which have a greater uh price

which have a greater uh price performance over the Nova 1 uh which is

performance over the Nova 1 uh which is another LM you can um you can consider

another LM you can um you can consider about it and this is only a part of the

about it and this is only a part of the new features and if you are free feel

new features and if you are free feel free to join it uh online it is free so

free to join it uh online it is free so just take a look on on that and register

just take a look on on that and register through the car code and also this slide

through the car code and also this slide I will send it out um after the sessions

I will send it out um after the sessions is and or this is for uh my today

is and or this is for uh my today presentations and sharings. Uh hope it

presentations and sharings. Uh hope it helps and hope the agentic IDE or

helps and hope the agentic IDE or agentic AI can help you to do the

agentic AI can help you to do the experimentations help you to validate

experimentations help you to validate ideas and then fasten the whole cycle on

ideas and then fasten the whole cycle on how you want to realize your business

how you want to realize your business value. Finally, good luck for your uh

value. Finally, good luck for your uh hackathons and uh yeah just let me know

hackathons and uh yeah just let me know if you have any questions etc. Okay,

if you have any questions etc. Okay, thank you. Thank you log. Um I spot a

thank you. Thank you log. Um I spot a question from the floor says, "Hi Mr.

question from the floor says, "Hi Mr. Yan to use Kirao. Do I need to download

Yan to use Kirao. Do I need to download it to my laptop?"

it to my laptop?" >> Oh yes, [laughter] we have in terms of

>> Oh yes, [laughter] we have in terms of ID. Yes. Uh because it's a desktop

ID. Yes. Uh because it's a desktop applications. So uh you need to download

applications. So uh you need to download uh the cur applications but you can just

uh the cur applications but you can just uh go to the this one this one. Yeah,

uh go to the this one this one. Yeah, this one. Yeah, you can easily register

this one. Yeah, you can easily register it and then you can download the

it and then you can download the application there.

application there. >> For participants, please feel free to

>> For participants, please feel free to scan the code and try

scan the code and try >> Kirao.

>> Kirao. >> Yeah.

>> Yeah. >> Okay. If you have any questions, please

>> Okay. If you have any questions, please type in the chat box.

type in the chat box. >> Yeah, I see. Oh, I see Edward had had

>> Yeah, I see. Oh, I see Edward had had questions, right?

questions, right? >> Yeah, Edward has just addressed

>> Yeah, Edward has just addressed >> Yeah.

>> Yeah. >> the questions as well. Thank you,

>> the questions as well. Thank you, Edward.

Edward. >> Yeah, thank you so much.

All right, I think um if it might take some time for them to type the question.

some time for them to type the question. So, we'll use this time um to thank our

So, we'll use this time um to thank our speakers again um as well as AWS for

speakers again um as well as AWS for being our technology partner.

being our technology partner. >> Thank you Log and Edward.

>> Thank you Log and Edward. >> Yeah, thank you so much.

>> Yeah, thank you so much. >> Sure. Thank you.

>> Sure. Thank you. >> Okay, we would like to share our house

>> Okay, we would like to share our house line now.

line now. participants, if you have any questions,

participants, if you have any questions, please type right now. Um, while we wait

please type right now. Um, while we wait for more questions, I would like to also

for more questions, I would like to also take this moment to thank our

take this moment to thank our co-organizers, Cyberport and Hong Kong

co-organizers, Cyberport and Hong Kong Science Park, and our government funding

Science Park, and our government funding organization, the Digital Policy Office,

organization, the Digital Policy Office, for their invaluable support. We would

for their invaluable support. We would also like to thank all of our sponsors,

also like to thank all of our sponsors, um, technology partners, media partners,

um, technology partners, media partners, and supporting organizations for their

and supporting organizations for their help.

Well, I don't see further questions at this time, so we will proceed to wrap

this time, so we will proceed to wrap up. If you think of any questions later,

up. If you think of any questions later, please don't hesitate to reach out.

please don't hesitate to reach out. We'll forward them to the speakers and

We'll forward them to the speakers and address them um as soon as possible. So,

address them um as soon as possible. So, the recording of today's session will be

the recording of today's session will be shared with all of you via uh through

shared with all of you via uh through email. Before we conclude, we would

email. Before we conclude, we would greatly appreciate your feedback. So

greatly appreciate your feedback. So please scan the QR code on your screen

please scan the QR code on your screen to complete the brief uh evaluation

to complete the brief uh evaluation survey. This would be greatly helpful

survey. This would be greatly helpful for us to improve our future workshops.

for us to improve our future workshops. I think that brings us to the end of

I think that brings us to the end of today's workshop. Thank you all for a

today's workshop. Thank you all for a great section. A friendly reminder, our

great section. A friendly reminder, our next workshop will be held on Tuesday,

next workshop will be held on Tuesday, 28th of January from 3 p.m. to 5:00 pm.

28th of January from 3 p.m. to 5:00 pm. The theme will be lower auditory um

The theme will be lower auditory um economy and IoT innovations powered by

economy and IoT innovations powered by HKT. So, please make sure to register in

HKT. So, please make sure to register in advance if you would like to join.

advance if you would like to join. Again, thank you so much for your active

Again, thank you so much for your active participation. We look forward to seeing

participation. We look forward to seeing you next week. Stay tuned and have a

you next week. Stay tuned and have a nice week ahead. Thank you.

nice week ahead. Thank you. Yeah. Good luck to everyone.

Yeah. Good luck to everyone. >> Yeah. Happy Friday.

>> Yeah. Happy Friday. >> Good luck. Hope

>> Good luck. Hope >> it's Thursday, [laughter] not Friday.

>> it's Thursday, [laughter] not Friday. >> Happy Thursday.

>> Happy Thursday. >> Thursday is in your Friday.

Videodaki o ana atlamak için herhangi bir metin veya zaman damgasına tıkla

Paylaş:

Transkriptlerin büyük çoğunluğu 5 saniyeden kısa sürede hazır

Tek Tıkla Kopyala125+ Dilİçerikte AraZaman Damgasına Atla

YouTube URL'sini Yapıştır

Tam transkripti almak için herhangi bir YouTube video bağlantısı gir

Transkriptlerin büyük çoğunluğu 5 saniyeden kısa sürede hazır

Chrome Uzantımızı Yükle

YouTube'dan ayrılmadan transkriptlere anında eriş. Chrome uzantımızı yükle ve izleme sayfasında tek tıkla herhangi bir videonun transkriptine ulaş.

Chrome'a Ekle — Ücretsiz

YouTube, Coursera, Udemy ve daha fazla eğitim platformuyla çalışır

Anında Transkript Al: Adres Çubuğundaki Alan Adını Değiştir!

YouTube

←

→

↻

https://www.youtube.com/watch?v=UF8uR6Z6KLc

YoutubeToText

←

→

↻

https://youtubetotext.net/watch?v=UF8uR6Z6KLc

YouTube TranskriptiSonuçların hazırlanıyor…

YouTube Transkripti:IOT Data Hackathon 2026 – Workshop 3