AI Agent for Information Retrieval
Materials — open to everyone, no sign-in
Topic: AI Agent for Information Retrieval
Presenter: Coach Ken
Additional Resources:
Description
Join us on Sunday, March 23, 2025, for a cutting-edge technical workshop titled “AI Agent for Information Retrieval,” presented by the renowned Coach Ken. This workshop, scheduled from 19:00-21:00 PST, is designed to delve into the fascinating world of Artificial Intelligence and its application in data mining and information retrieval.
This interactive session will take you on a journey through the latest AI techniques, demonstrating how AI agents help in the effective and efficient retrieval of information from diverse data sources. Attendees will gain a deep understanding of how AI can streamline data access and enhance decision-making processes in various sectors. The workshop is a must-attend for tech enthusiasts, data scientists, IT professionals, and students who want to stay ahead in the rapidly advancing field of AI.
The value of this workshop lies in its emphasis on practical, real-world applications of AI for information retrieval. Under the expert guidance of Coach Ken, you will not only learn the theoretical aspects but also engage in hands-on exercises. Don’t miss this opportunity to refine your skills and broaden your understanding of AI’s power and potential.
Video Transcript
all right we can get started uh it’s great to see everyone in this event again um so uh today we are going to uh talk about uh a we we are going to show you a prototype on how to retrieve large amount of information uh but let’s get started with the AI front uh as everybody know that um AI is everywhere now and in a lot of software um we see that the companies are adding features that are AI enabled this help uh users to interact with the software and the system uh sometimes it’s Hardware uh more naturally uh they can use their natural language instead of um using the mouse and keyboard uh and they can get information also back from uh the the software using natural language uh and some things that are better for human interaction um so this may include uh chatbot um like support sales uh and include some uh other features uh like uh creating slides writing code uh a lot of things are uh a lot of software uh is adding the AI capabilities within them uh to make it easier for the users to use right so uh with this uh advancement in AI uh we uh the customers the users expectation is also increasing uh they feel that they can rely on this uh AI to retrieve information that are fresh uh that are vast and they may want to uh in some cases retrieve information that is uh private like their own uh documents that are within the Enterprise or in on their computers uh however uh the uh chatbots that we been so far had uh challenges with retrieving uh fresh information and private information uh because they underlying they use large language model uh which is trained on large data set uh but it’s trained usually in a time in the past uh so if we want the uh large language model to return information that is up to dat uh usually that is uh challenging because they have not seen that information in their during their training uh so the answers they provide uh would be out of date or they may hallucinate and come up with information that is not accurate um to solve this problem uh people come up with the concept of uh retrieval augmented Generation Um which means we uh before we query the large language model we actually use some type of search engine to do some research ahead of um ahead of time uh based on the users’s query uh and then uh we get the results from our own database or our own search engine um and then we use the results as part of the prompt to the large language model uh and then the large language model can understand oh here are the relevant results and here is the question that is uh asked uh on these results so based on that input much more enhanced upto-date input they will be able to provide better answers uh than just relying on their own information that is that may be stale uh so that’s the way that um uh usually people these days how try to solve the the the problem of retrieving fresh information uh using the chatbot um so last weekend uh we had coach Cindy to give us a talk about how to build a retrieval augmented generation system uh and roughly what the architecture uh may look like and here is uh so if people are interested uh you can so we are going to re-upload the uh the talk again on YouTube soon uh if you missed last week uh information you can subscribe to our Channel and then once we upload you’ll get the notification uh for the uh for the talk last time um so here’s one of the diagram that uh Cindy used uh to illustrate how we deal with uh large amount of information so uh we would uh try to break the information into chunks um and then uh and then we will do some index uh on the on the document and then when the user try to query the information um we will try to find the relevant chunks uh to answer those questions uh and then combine that with the uh with the users query and pass that to the large language model the reason we have to do that kind of work to uh index the documents is because uh sometimes the documents can be really really large or can be there can be lots of documents so um the large language model will not be able to they they have Li Li mited context window when you ask a question to a large language model uh it can only take up it can only consume or um accept uh so many characters or so many tokens uh so they there’s a upper limit on how much the large language model can accept so we cannot feed everything that we have which may be private database which may have lots of information into a single query to the large language model um that will not fit their context window so they will not be able to you know process so much information in one big uh big question uh so instead we try to find the relevant chunks uh to uh feed to the uh llm uh and that size could be much smaller so before we ask the large language model to answer the question we actually do a query on our own using some search engine a vector um these days we use vector database uh to find the relevant text um that are some paragraphs which I will demo later some a few paragraphs that are relevant to answer users’s question and then we combine those paragraphs plus users question into a more complete uh question to ask the life model to summarize uh based on the information that we retrieved uh so this is a talk that Cindy went into more depth last week and we are going to post uh video again uh on YouTube so you can subscribe to our channel uh but today we are uh I uh will demo a prototype that is based on that idea uh and and try to uh help us to understand like a book that is really really long um so part of technique uh with these document indexing uh and uh uh especially very large chunk uh very large amount of data uh is we we may not want uh very small paragraphs some some of the paragraphs may be very local so you you may ask a question like oh where did the this person go um on this date right so so for very specific questions um uh the small paragraphs will be able to answer that accurately if the person and the date you know appear on the paragraph uh but uh if we want some kind of more higher level questions like um uh what are the uh what are the unique features of this service um that may go across multiple documents or may go across multiple paragraphs and um very paragraphs may be very specific to one particular Point um so uh when we retrieve uh that we may actually miss the big picture um so if we want some uh to answer some big picture questions uh sometimes we have the technique of um aggregating smaller paragraphs a number of smaller paragraphs into a summary and then index the summary as well and then we have multiple summary chunks and then we also uh combine those summaries into even higher level summaries to uh so that we are able to uh have some paragraphs that are uh covers uh more breath uh than very specific points uh so so that uh in reality will become some kind of hierarchical index uh that not only index the lowest level of detail but also index higher level uh constructs so we combined uh uh documents into summaries and then summaries into even uh more abstract summaries so we can answer more abstract questions so we’ll try to demo that um so uh the rag alone is not going to give us the user interface uh to actually ask these questions so to ask these questions we still need some type of you know chat interface or search interface which will allow the user to type in uh our question and then uh show the result uh so uh that goes into the realm of AI agent which we had you know previous events talk talking about AI agents uh AI agents can take input uh both from the user as well as the real world and then you will try to process this information and then it will generate some results sometimes it’s just um answers to users questions sometimes they may be able to perform actions on users behalf such as you know planning the trip buying a ticket making a phone call and so forth so this General framework of AI agent can be used uh also for information retrieval um and the input part uh is in addition to users input we also provide uh the uh the the uh R you know the the search results as part of the input to the AI agent for it to process and then um provide us a summary of what we want want to know um so our demo today will be on Discord uh so I have hooked up an AI agent uh on the front end using the Discord uh chatboard interface so it can listen to a chatboard uh the Discord Channel um and also answer to Discord and then on the back end of the uh of our agent uh I hooked up with a uh retrieval augmented generation system uh with uh the hierarchal index and Vector database uh to be able to search the information that we put into the index um so you some of you may have seen the uh some may some of you may have seen the video that I used to introduce this event uh so uh we are going to for for just for demo purpose in this event we’re going to help you to retrieve information uh for a old novel um and uh it’s a novel called Mo mck um it’s a very uh famous American novel from the 19th century uh about a uh a crew of Whalers uh so they they catch whale for living and uh uh so so that there is this story of a captain uh who who want to chase a whale and he chased it across the ocean and then he got into some disaster because the the force of nature is just too huge for uh humankind uh so there’s this uh famous story uh that uh it’s in the novel form and um it has 135 chapters uh it’s very popular uh there is um this a Gutenberg project where you can find a lot of uh classic novels and books you can download and it’s U constantly this novel Mo Mobic is on the you know top 10 downloaded uh books list so I use that to illustrate you know how we can use our latest AI technology to quickly understand a book and uh try to understand some details of the book without uh needing to go through all you know 135 chapters uh to uh to to to uh to read through and then get the get the information because um these days all of us are busy so uh we may not have uh as much time so highly recommend it you spend the time to read it but if you don’t uh we can use the uh rag process to uh quickly extract some information from the book and so this is the Discord channel uh we are I’m going to also type the Discord Channel link to the YouTube yeah uh we are going to type the the the link to the uh YouTube uh chat uh so that we can uh uh you can join the Discord Channel all [Music] right okay uh there’s an audience uh person said there’s no sound uh are people able to hear me now uh hold on hold on one second all right uh so we have I have sent the uh link to the to our Discord Channel and uh here is the uh user interface of Discord once you join our Channel you can see uh uh yeah once you you join the Discord server uh you will see multiple channels within you know our our Discord server uh so uh on the left hand side we see multiple ones uh we can click the novel uh so the chatbot will actually respond uh to the novel channel uh in terms of the query uh that we that we have um so I have been already been testing this uh bot the bot is running uh on my IDE uh so uh I used so I used cursor uh which is AI enabled uh integrated development environment to write code uh and I have uh coded the uh the rag system uh as well as the uh AI agent the chatbot system um so now it’s all hooked up together uh can uh briefly go over what the uh how I did the index and the uh retrieval augmented generation system um so uh we have this novel uh which is the mobii the F text uh now it’s uh it’s open source it’s available for uh download um from Gutenberg project um you can also download other classic uh books from there uh and then uh once I have this um novel this is a very very very long novel it has 200 about like 200,000 words uh so uh it takes a very long time to uh to read through if um we just read it by ourselves um and then uh I break it down into multiple levels um the uh the first level is essentially just uh uh cutting the novel uh into chunks uh based on some character limit and then each chunk have a slight overlap with the next chunk uh so so each chunk there are like 2,000 words and then another 2,000 words but with like about 500 Words overlapping so uh so I chunk it up uh to level one uh level one is essentially the original text of the author um and then from level two level two is a uh so if you can see level one um yeah so this is kind of the original original text from the author um and then level two is the uh summary of level one data um so every chunk becomes shorter uh using llm uh so instead of 2,000 words for each uh each Chunk in level one now it becomes like 200 words uh in level two so level two has as many chunks as a level one um but it’s shorter uh and then level three um is even shorter so each one would uh let’s see yeah each one would combine uh multiple chunks from level two like about I think it’s about like uh five yeah something like between five to 10 chunks of level two uh and then summarize that into a another chunk um so every level there are fewer chunks and smaller shorter and uh and a more uh brief or concise summary of the previous level so the level one is the most detailed including the original text level two is already summary level three is summary of level two and then we in the end we have a single summary of the uh the the whole book uh so let’s see this one yeah so this one um is just based on the large language model it’s a uh uh if we were to write the whole novel into a single article abbreviated version of the a bridge version of the novel then here is what it will look like um so there’s this n narrator called ishmail and he’s a sailor and then uh he went to a ship um uh and uh let’s see uh chapel and okay so so this are setting up the setting up background of uh the the environment where uh this uh this sailor grew up in uh and and then uh they uh the Sailor boarded a ship called uh peod um and uh and then he met uh with uh the some captains and eventually uh he would uh travel with this Captain Ahab to uh the ocean and then this is a w wailing so the waing industry uh uh was there in the 19th century and there are lots of you know people uh that make a living by catching the whales so he boarded this uh uh this ship uh with this this Captain um and then uh eventually eventually they Tred to uh catch the whale but then uh the the whole ship sunk uh because the um they were able to catch the whale but the whale also knock out the ship and then they all kind of sunk into the sunk into ocean and then there were some uh some Journey like some summary and then uh in the end the only the Storyteller survived this uh this disaster and then he was picked up uh by another uh by another crew another ship so that’s uh if we were to write the whole novel into a a single article then this is the this is article that uh we are we are going to we are going to write um and then uh with that uh we can ask using the uh using the agent interface or the Discord uh chatbot uh we can ask some questions about the about the novel uh and and when we ask these questions uh when uh the AI is going into different levels of abstraction of this novel um to see which type of summary is most relevant uh to answer my question um uh and then you’ll prepare some you know list of a list of relevant paragraphs some are summar some are original text um and then uh it’s going to write a short essay uh uh initial so so the essay is going to be written in the tone of the original author uh as if the author is going to extend his novel or he’s going to write a bridge version of this novel uh in this uh in this new environment uh so uh to make that uh more interesting so instead of very robotic response um the uh it would actually respond with the author’s writing style as if it’s part of the novel itself um uh so so now if you’re on the Discord Channel you can go to this novel uh yeah this novel Channel and then you’ll see me uh interacting with the the chatbot uh to ask some questions about the about the novel so let’s first ask about the the crew right the crew uh the crew and their names okay so it’s processing uh as I mentioned yeah it went to it went to different levels of detail trying to uh retrieve supporting uh supporting material to answer my question like the crew and their names uh and then with the supporting material uh it it writes a small essay as if it’s part of the book so if you ask original author to write it it will be kind of like this yeah the crew of uh po po cord was a diverse and uh temp temp tempet to slot uh and then uh each one with uh with a tail to tail and then it actually listed uh a bunch of people on the on the on the ship uh captain aab and then there was uh stop uh and then there’s this uh build dead and Pac uh and then yeah so there are a bunch of people and then there Starbuck flask uh so so each one uh it described uh what they what they are like their their background um and then uh uh what they do right so so it’s almost like a it’s almost like a job interview where you need to do the you know the situation and then the action and the result right so the yeah so this is the crew they want to catch whale um and then uh the action they took was to go on a trip to hunt the whale uh and then the the result uh is uh they actually got into some kind of uh Mutiny uh which means some uh some crew actually disobeyed the captain uh and then the captain was not able to maintain control of the uh uh of the ship um and then let’s see the crew bur to the danger of the sea Voyage yeah uh uh so so so that’s sort of the uh the the result is the captain actually uh was really kind of motivated to catch the whale and he is kind of like a dictator but the the crew has their own personality and they at some point stop listening to the to the captain and then they had a mutiny so they kind of like a revolt and rebel Rebellion on the on the ship uh so so that’s the that’s the kind of a result uh so so this is the um in response to the crew and their names um uhy uh if we scroll down more um I have asked the chatboard or I quoted the chatboard to print what are the supporting evidence that you actually retrieved from uh our database so our database remember is multiple level level one is the most detailed level uh level two is a is a summary of level one and a level three is a summary of level two so it’s like a hierarchal index all the way up um and then uh in terms of answering this question the crew and their names uh it actually retrieved multiple paragraphs um some in it looks like the first one is in level two uh and and there are some this is already a summary level and then there are some that are from level one which is the original text of the uh of the from the author um and then there is this you know one is uh the whale ship prepares for voyage the other one is stop is one of a car character in the in the novel uh and then there’s this Mutiny at the sea uh so essentially the there’s a Revolt uh from the crew uh and and they disobey the the captain um and this is also level one so in in terms of a hierarchical index is the the base level um uh from the from the original author uh and then uh the fourth one is uh this person he has prepared for another Voyage uh this is level two it’s a summary of level one so is not original text and then this one chaotic whale hunt uh is the the final the final paragraph that’s used as supporting evidence to answer my question uh again this is level two um so you can see the the content uh where the rack system uh actually retrieve from the vector database uh to answer the question um you can also see the uh the the number one the level and then number two the the the origin the text that is inside that chunk inside that node imagine a tree of information where the lowest level is most detailed and then there summary on top and a summary on top right so you can see actually what is stored in that note um there is also this interesting distance um calculation uh which is U um based on my question uh it uh find that it tried to map my question into a point in a multi-dimensional uh space um and then it try to also map all the three nodes into this multi-dimensional space um and then you try to find the five closest point to my question um and then this is the distance within that Vector space so um the closest one is this one um I guess about the uh let’s see the the closest one is this level two uh the distance is 1.22 and then this so you can see distance field uh gradually increasing so these are the five points that are closest to my question of uh uh This original the crew and their names right um so so that’s how uh using the knowledge that I got you know from last talk from coach Cindy uh I developed this Rec system uh a sort of prototype miniature system system uh that actually do this hierarchical index and then uh try to uh answer the question uh that the user provides to uh and then find the relevant text to to prepare an answer and then that’s the that’s the answer in the author’s original home so we can ask some other questions like uh is whale dangerous yeah so back I guess back in 19th century our navigation technology is not that advanced and um uh at the high seas is a there’s a lot of danger uh trying to catch whales um so uh so here uh it had a summary of uh actually write he wrote a small essay like the original author uh to summarize um the danger of uh wailing industry or the The Wailing Journey um so uh there’s this uh about the sperm whale uh it’s very dangerous um inside the book uh it actually is kind of partially fictional partially historical so it actually quoted some uh detailed like uh history of uh Seaborn disaster where uh during catching whale the the ships actually um was sunk by whale and uh uh uh so so so that was extracted from the text and then was summarized by this uh you know the the rack engine the retrieval augmented generation engine uh and then uh there is a person Commodore J something uh that is uh not is skeptical of these Tales um but uh there are more supporting evidence that uh uh the The Wailing uh on the ship is quite dangerous the whale actually may sink ships and actually they they sunk some ships um so this uh is again based on my question is whale dangerous and then um it actually found some paragraphs on both level two the summary level and level one original text they are related to the danger of wailing um and and this one uh first paragraph has very specific example like Egypt’s uh plagues oh yeah the specific example is 1820 there’s this particular uh ship that is sunk by the whale um and then uh uh there this uh the whales power and historical attacks and then uh ocean power um all these are uh I guess because I say the whale and danger is whale dangerous um it actually found very uh some paragraphs that has very small distance to my question so as you can see uh this distance 0.7 and 0.7 0.8 about 0.8 uh they are actually closer than the the previous question so for the previous question there they were not able to find very concise paragraph that match my question so my previous question was uh tell me about the crew and their names right so distance here the closest for that question was 1.2 2 two um but um uh uh so so they had to gather so for the previous question they had to gather different paragraphs of about different crew members and then combine them together uh but but but about the danger of whale they actually found very specific paragraphs that are really talking about the danger of whales uh and the distance is much closer you see the 0.7 right it’s almost like half as close as the uh uh when they prepar the answer for my previous this question um yeah so uh so that actually shows if you are interested you can actually I think you can just type questions here and it will as long as my uh bot continues to run it will answer the question uh of the the users uh so you can just type into this channel um okay more about uh okay yeah some user is typing okay so we can yeah uh so if you have time you can actually play with this uh prototype uh okay what’s the most critical plot point in this book this is kind of uh abstract but we can see okay that’s submitted by a user uh by by an audience member I guess title the vengeful hunt and bongs of the sea and uh uh p packe is uh the ship name uh and then there’s this uh mobid dick is the the whale the the name of the whale is mck there’s a powerful whale that uh the the captain is obsessed with uh so so that’s the uh let’s see but was not the vessel a piece and then unlikeness yeah so so that is uh so of the answer they provided um con w yeah I’m trying to understand uh it’s kind of like Old Style English so it’s not too it’s a little bit more difficult for me to quickly scan through and understand it yeah consum his path Hunters the crew hunters and dreamer were join into the F title yet Remain the Beacon of Hope okay Humanity could still face okay so it looks like the the critical plot is related to the danger of the the whales and the humanity so uh I guess two parts one is the humanity against the nature right human against nature right so this one crucial plot another plot is the uh there’s this um this crew member uh who is very queen queen quack uh this person seems like uh very stable uh like there’s with the Storyteller Storyteller is the is ishma and then qu qu um and and their friendship seem to be the Beacon of Hope in terms of this uh uh this storm in the heart of the this storm that is human versus nature all right so so that’s kind of interesting and you can see the uh the distance they are not very close but it’s better than the list of Crews but so they found a few supporting evidence on uh uh on this question yeah so uh so that is what we are going to demo today for the uh yeah for for reading the book using the rack system uh in summary uh we pre-processed the book into a hierarchical index uh of multiple levels and then uh in the runtime when we receive a query from the user we go to calculate distance between the user’s question and all these uh uh all these notes in the hierarchical index and then find the closest notes five noes that closest and then use that as supporting material for the large langage model to answer the users’s question and then we uh eventually use again large language model to summarize it in the tone or the writing style of of the original author and then compose an essay as if it’s part of the the novel yeah so the technology uh you can imagine uh is not just applicable to you know reading this book um if you uh if we have a lot of data like uh private uh document design documents uh product requirements uh you know uh the the stock reports um if we have lots L lots and lots of data uh that uh that are tax based uh that we want to uh quickly summarize and uh answer some of our questions then this is a potential technology the rack retrieval augmented generation can be used to help us uh to uh do summary and uh answer questions on large amount of text that the large language model has not seen before all right so let me go back to the presentation okay yeah so uh we demo that uh the rag system in the context of AI agent and uh eventually uh it showed up as a a chatbot within the Discord um and so uh how uh so so this system is uh actually implemented by lots of companies uh when they need to answer customers questions automatically so sometimes when the user goes to the website or the app of a company uh that may be you know doing travel or doing financial planning they may want to ask some questions related to the product that they offer um and then the uh different software companies actually deploy or implement this retrieval augmented generation system in the back end and then show show up as a it shows up as a chat agent uh on the on the web page um to answer customers questions um and this is uh quite useful uh in many uh situ situations when we actually need to introduce AI enable features to the to the customers uh so how do we actually go about building an AI agent uh then uh we will uh how do we start right uh usually we can start from the very beginning right I have a a chatbot that has small amount of capability and then we gradually add more capabilities uh to our agent to make it more full flag um so now uh my assistant JC is going to demo how we can actually build an AI agent from scratch all right hello can you hear me yes I will start screen sharing if you turn on sharing for me so here’s a request is it visible uh it’s black on my side okay so now yes now it’s visible all right so in our class which we I’ll be demoing is how we will build a chat bot by using a few things firstly with a cursor’s code editor which is a powerful AI tool editor which allows us to make code a list of requirements which gives the AI tool editor uh instructions on how to make a chatbot how we generate code and how we run it it so we’ll be building a pretty simple demo chat boot for you to see and it’s all inside of this cursor’s AI ID which is an extension of Visual Studio code which allows it to connect with Claude 3.5 in order to write code pretty fast uh as the project we’re about to make is reasonably simple which is to make a chat bot uh the actual prompt itself can just go through a single list of requirements and allow the AI to make the entire thing which works in a single prompts and a single run to create an entire thing for more complex projects it’ll require more than just this but as I’m just demoing this for now I will show you it only takes one run so to do this we have cursor’s IDE which has has a file system manager on the left in the center a list of contents of the individual file uh and on the right a chat area which allows communicating with a chatbot which can which will help us write the code so I’ve already prepared a list of requirements beforehand which asks for a few things in our chat box firstly I wanted to be able to process and respond to user input I wanted to be able to support natural Lango language queries I want it to use an open AI API key and I wanted to build the chat bot in this area I’ve provided an example call to open AI with it as this is the most upto-date version of it so it’ll connect correctly with chat gbt now on the right I have the list of files which I can add as context for the uh AI for clouds to see I’ve added this requirements file on as a context and I’ll be asking it simply to based on the requirements in this requirs file build a chatbot as you can see the chatbot will now create an entire functional system of code which allows it to communicate with chat gbt for a reasonably simple chat Bo it’s should be running we apply and accept these changes as you can see the cloud has already written up a full chat out for us it goes through a few things like an initialization a few responses and a main system which we will soon see demonstrated uh on the bottom of this IDE is a place where I can run code directly so I’ve already navigated to the folder we’re in and I’ll just do python name y although uh cursor AI can run with different languages I’ve chosen to use Python because it’s the simplest and prevents the most is strange errors from occurring so as you can see there is a chat bot and I can ask anything what’s the best type of cracker this chat bot is directly linked to open ai’s chat GT model so it can talk with the same strength as Chach although it doesn’t have the browser based search functionality which open AI you can find on the browser it’s just a simple call to the API and it allows it to do this next I’ll show you a slightly more Dem a slightly more advanced version of this which allows it to access a few more apis in the process the system is primarily for a traveler who is trying to go around the world and understand what the situation of their destination is and it allows this to do it through a few apis firstly a weather AI API which connects to open weather maap and a get exchange rate which allows it to find the exchange rate between any two currencies if you’re a traveler coming from the US to for example uh India or China it’ll be able to find the currency and the rate in between the two I program this entire thing within two hours or so maybe 1 hour using the AI and it’s pretty powerful this thing has two main API functionalities although inherently it’s just based on chat GPT and allows calculation of intents as in what the user wants from the apis for example I can ask you something simple like uh what type of tower is the most famous in London as it’s just linked up directly to chbt it can respond to these with anything the most famous tower in London is the Tower of London also known as Her Majesty’s Royal Palace but with the API it can get for example live weather in London which I’ve had the AI code for me like what’s the WEA WEA in London and it can be written in any way like weather in Moscow now if the audience has any questions for what cities you want to know the weather of I can add do that but I will first uh and you can send requests for the cities in the YouTube chat screen for exchange rates I can also do this by directly asking convert 100 yen to CNY no Euro for example and it’ll immediately do that rate and we can confirm that this is around accurate although it’s a little bit off sync with the current exact one 100 to suggest a city uh you can type that into the YouTube chat window as you can see it can also convert between any two currencies and if you have any questions on that you can immediately show you uh I can also ask 10,000 uh c y to rupes they can go in between any type of C see to show you this how about Canadian dollar versus US dollar uh one Canadian dollar to USD does Canadian dollar have an abbreviation CAD CAD yes it shows the exact rate as well although it just gave me what was the General conversion on the side okay I guess it rounds to the nearest scent because that makes sense it’s great yeah $ 100,000 Canadian dollars equal to 69,000 us any suggests from the audience uh not at the moment all right this will be open um so this is what I’ve shown for these two demos which were both this first one being coded within 20 seconds and the second one being coded in half an hour to an hour most of which was spent setting up the API Keys thank you Ken I will hand it back to you so people have questions for JC on which cities or which currency pairs feel free to type into uh into the chat window all right so let me take over the presentation again okay yeah so uh so first uh we can quickly summarize what JC did uh essentially we uh built a simple AI agent at the beginning and then uh add it a couple features uh such as checking the weather and converting between currency payers to the agent uh we uh use the AI cursor editor uh to help us code um and to communicate our requirement to the cursor uh we write a document almost uh as if we are working in an engineering team uh the reason we use a document is it’s sometimes very difficult to communicate uh the idea through a single sentence like we want to do this so um in a real you know software company we write requirement docs so that’s how we actually are able to be more specific about what we want from the code uh and then we use the requirement uh and ask the cursor to generate code for us and then we run through uh run through the code and then gradually we added more and more features um yeah uh right now JC’s demo is running through the terminal uh but in real life after we build that step we are probably going to build uh the steps to show the uh show the user interface uh in on the web or in mobile apps uh so that leads to our upcoming AI courses uh so today we uh have limited time so we can only demo so much uh but uh to help uh help you learn uh about these two technologies uh the retrieval augmented generation uh as well as building the AI agent we have prepared courses on those subjects uh so the first one is uh the AI agent for beginners um and uh I will be teaching this course uh to uh over five evenings uh each one with one hour uh and the goal is uh one um I will help you to go through how to uh create an AI agent uh how to uh how to let it collect uh real world information and uh how to process it uh using different um you know thinking patterns uh and also how to perform actions uh responding to the user or performing more actions on behalf of the user um so the target audience of this uh is be beginner developers uh you could be student you could be you know entry level developers uh it’s very uh beginner friendly uh and uh this course will start uh at the end of March uh so yeah so I have the QR code to join this course here and uh so so this is one of the courses uh yeah so my my own background uh I have worked in the software industry for uh over 20 years and recently I started mow school and I’m full-time doing this a coaching Ser coaching service for we provide coaching for uh software professionals including developers uh engineering managers uh leaders to uh be more more successful in their software career uh plan their career uh help them with transitions uh either uh doing interview or um getting promoted or uh getting more responsibilities such as you know becoming the first Frontline manager uh in the uh for the first time uh so uh we have been uh providing this service to hundreds of students uh and we have been successful you know helping these students uh find their next opportunity and expand their scope um yeah and then so this course I will teach this course and JC will be my assistant uh and this will be uh targeting beginner developers uh with you know basic coding skill uh and by the end of this course one is you will follow if you follow through our schedule you’ll be able to come up with a chatboard uh with multiple features um second is uh you can have your own personal project to build a AI agent that uh whether you want to build a support agent or travel whatever you want to build you have the opportunity that to um follow the guideline outlined by this course and then build your own AI agent as well so this course will start uh uh this Saturday so coming Saturday uh March 29th yeah and then uh there is another class uh which is on retrieval augmented generation uh which is the part that I show you earlier uh tonight uh how to retrieve uh information from large amount of data uh and the target audience uh will be software professionals uh who are familiar with you know software engineering uh but they want to get into the field of AI um and machine learning uh and they want uh to gain some skill uh in in this area uh in terms of just breaking into you know the machine learning AI field um so this will be taught by uh coach Cindy uh who made a presentation last weekend and so she is a uh director of uh machine learning uh frontend engineering and data science and she has lots of experience building AI systems um and this will be uh over four even ings and she will cover uh Advanced topics in uh retrieval augmented generation um yeah so the first uh cohort uh will start on uh Friday April 4th uh so maybe about a couple weeks from now uh and uh yeah so you are welcome to join our uh rag course as well so here is the QR code to join it all right and uh so uh if you just uh you are interested uh feel free to join our Linkin group uh as well as our WeChat if you’re using WeChat you can join our WeChat group as well and then I have left these two uh QR code for our upcoming courses here um I will also type the URL from the uh from the YouTube chat uh for you to look at our courses uh so we host uh lots of events uh if you want to subscribe to our uh YouTube channel you’ll get notification for future uh events and we also share knowledge uh through YouTube and Tik Tok um to to help our community to do better you know in the in the software industry so here is the link all right uh let’s see uh does anyone have have any questions I’m happy to answer any questions yeah thank you for you know joining our uh our meeting today our our event today all right okay so um if there are no questions uh yeah I will leave this up for a while uh feel free to join our groups uh and our courses um and uh yeah we hope you can become more successful uh in terms of developing software and getting into the field of machine learning and AI all right thank you everyone for joining our event e e e e