# 01 2026-05-10 Introduction to Generative AI LLMs

course: Module 4 — Generative AI & LLMs
module: Module-4-Generative-AI-LLMs
date: 2026-05-10
type: transcript
video_url: https://personal-learn.armco.dev/files/_Recordings/Module-4-Generative-AI-LLMs/01_2026-05-10_Introduction_to_Generative_AI_LLMs.mp4

---
[00:16:35] Google, by the way. Right. So Opal N8n
[00:16:39] And I will be talking about something very interesting today, which is called
[00:16:45] Agentic development environment
[00:16:49] Right
[00:16:50] So, all of us, I'm assuming since all of us become
[00:16:56] A programming background. All of us, we know what is IDE
[00:17:02] Isn't it? Integrated development environment, right? All your PyCharm, Spiderlime, they're all examples of
[00:17:14] IDE, integrated development environment, right?
[00:17:18] No.
[00:17:20] This IDE is something which
[00:17:25] has now improved
[00:17:28] Right into something known as
[00:17:31] So, earlier what we used to have, or even now we have it, it's not something which is obsolete, but earlier what we used to have is
[00:17:41] What we used to have was called Integrated
[00:17:49] Integrated
[00:17:52] Development
[00:17:59] Environment
[00:18:07] Right.
[00:18:09] Examples being VS Code, PyCharm, Sublime, everything. Right
[00:18:16] What has happened is we have started integrating agents, okay, code completion kind of agents into these ideas, making them
[00:18:28] ADEs. Okay. Agentic development environment, and that's where, let me just first write it down. This is
[00:18:38] Agentic.
[00:18:41] Development
[00:18:47] environment.
[00:18:56] Okay.
[00:18:58] Agentic development environment, and this is where
[00:19:02] A lot of new, I can say, ADEs coming into picture. For example, you would be using or you would have heard about something known as cursor
[00:19:15] You would have heard about something known as the replit.
[00:19:23] Right, a few of you would also be using it
[00:19:26] Same way, there is another one called anti-gravity from Google
[00:19:33] This is Scott
[00:19:35] anti-gravity
[00:19:39] Correct.
[00:19:41] So these are all I can say different different ADEs Agentic development environments, right? And these ADEs, they help us in
[00:19:55] prototyping the output very quickly, right? These agentic development environment, they help us in getting the output
[00:20:04] in no time, and the good part is that you don't explicitly have to write a code. It is the agent which will write the code and will get the things done.
[00:20:17] Now the agent, it might use Claude for getting it done. It might use Gemini for getting it done. It might use, you know, GPT models for getting it done. That is a different discussion altogether
[00:20:32] what you see in front of your screen is basically the way how things they are looking like at present.
[00:20:42] Okay
[00:20:43] So, in the previous session, I have already given you some experience
[00:20:49] With N8n, right? Today we will first start with Popal right? I will spend some time
[00:20:58] explaining you how can you work with Opal and then we will directly jump into ADE also
[00:21:06] Specifically in ADE, I will talk about cursor
[00:21:11] Right. If time permits, I will also talk about anti-gravity
[00:21:17] But my first and foremost priority would be cursor, right?
[00:21:22] So this is the agenda that we are going to experience today. Right
[00:21:29] No.
[00:21:32] In order to understand OPL
[00:21:34] You will have to, first of all, understand
[00:21:38] How is it different from N8n?
[00:21:42] Right. In the last session, I'm assuming all of us, we were able to create those agents that we were trying to develop. And you know how the interface of N8n as a platform looks like. You know how the canvas
[00:21:57] it looks like, right? Now, when it comes to Opal, let me just first show it to you, and then we will discuss further. Let me just search for Google
[00:22:09] open. This is right now in its beta stage, right? Google Opal right now is in its beta stage, but it has been made available to us. So we can anyways
[00:22:25] And that's why it says opal experiment, right? So, what I will do is I'll just go to Google, I'll search for Opal. I'm sharing this link also in the chat box in case you want to
[00:22:42] Access it from here, right? So
[00:22:45] This is the platform, Opal. You just have to create your account, I mean, just Google sign in, and the moment you are able to do it
[00:22:55] On the right-hand side corner here
[00:22:59] You will find there is this option called
[00:23:02] Create new
[00:23:04] Now, the moment you click on Create New
[00:23:07] It will take you to a platform like this
[00:23:12] And this is how Opal it looks like.
[00:23:16] Okay, this is how open it looks like. On the left hand side of your screen which occupies more than three fourth of your overall screen, you will be able to ideate, you will be able to
[00:23:32] Create your workflow, right? In the last session also, I was trying to emphasize a lot on workflow, right? So that is something which you'll be able to do on the left-hand side of your screen.
[00:23:48] And on the right-hand side of your screen, you will be able to see the working prototype. Now this is something which was not very much possible in case of NA10. I mean, you could see the output
[00:24:03] Right. But you were not able to actually play around with it
[00:24:08] As if it was a application, right? So that part was slightly missing in nature. Here, it is not that case. Here, you can very easily try and build something for yourself, and most importantly, you can see that in action
[00:24:25] You can see that in
[00:24:28] Okay
[00:24:29] So
[00:24:31] This is what we will try and make ourselves understand, right? Now, how does that work and how we are able to experience this is something that I will discuss as we are progressing
[00:24:46] Right? So, again, for different, different departments, for different, different
[00:24:54] You can always make use of this. So what I will do first of all is I will try to create
[00:25:06] And HR
[00:25:08] As you may scanner agent
[00:25:13] Okay.
[00:25:15] So I'm creating an HR resume scanner agent, right? Usually, what happens is when you take a look at, you know, the traditional way of working, you would notice that the HR professionals, what do they do?
[00:25:33] They have a job description
[00:25:36] Great.
[00:25:37] They have a job description
[00:25:40] They have, let's say, the resume
[00:25:44] Of a candidate
[00:25:47] Okay.
[00:25:48] As you may offer
[00:25:50] candidate
[00:25:53] Okay.
[00:25:55] The job description will have have couple of
[00:25:59] Keywords
[00:26:02] Isn't it?
[00:26:04] The resume should also have similar
[00:26:09] Keywords
[00:26:10] So that the overall ATS score, as we say, right, that ATS score is decent enough. So there will be keywords in the job description, there will be keywords in the resume of the candidate, and what we will do is
[00:26:27] We will try to do a matching, right? We will try to do a matching. So what we will get as an output is
[00:26:36] Match score
[00:26:43] Okay, we'll get a match score. Now, what is this match score?
[00:26:50] Math score is, let's say, something which is between 0 to 100, and this score will be able to tell me that yes, whatever candidate I'm trying to screen
[00:27:05] Whether this candidate is eligible for the next round of interview or whether this candidate is not eligible for the next round of interview
[00:27:15] Right. So, you have job description, you have a resume of the candidate
[00:27:21] basis that you try to find out the match score. Match score is nothing but the keyword overlap, basically, right? It is nothing but the keyword overlap
[00:27:34] And if the match score is good, if the match score is
[00:27:40] More than 70, 80%, then maybe what we can do is, we can recommend this particular candidate, right? We can recommend this particular candidate for the next round of interview
[00:27:55] Right. For the next round of interview. So that kind of a thing as of now
[00:28:01] But multiple capacities, right? We are going to design a workflow which will try to automate this as a process. So I will have the match score. Okay
[00:28:17] And basis the match score, I will give my final recommendation. Now, when I say I will give, it is not me who will do it. It is my assistant, it is my agentic workflow which will
[00:28:32] do this. So it will give the next set of
[00:28:36] Recommendation
[00:28:43] Okay.
[00:28:45] So, job description, resume of the candidate, match score and final recommendation. Okay, now that recommendation can be that, you know, this candidate is not a good fit. The recommendation can be that, yes, this candidate is a good fit, and maybe we can consider his or her profile
[00:29:02] For the next round of interview, whatever, right? So this kind of a workflow is something that we are going to design on
[00:29:13] Let's say Google Opal
[00:29:16] Great
[00:29:17] How do we do it? Let's come back.
[00:29:20] How do we do it?
[00:29:23] When you are here in your Google Open
[00:29:29] You will be basically playing around with three of these nodes
[00:29:34] You will play around with user input
[00:29:37] Okay. You will be playing around with generate node
[00:29:43] And you will be playing around with the output node. Okay, so you will have user input, you will have generate, and you will have output
[00:29:53] Great.
[00:29:54] Now let's try and understand
[00:29:56] What should be the user input for you
[00:29:59] In this kind of a workflow, right? And if I just try and take a snap of it, and maybe if I bring it back to my blackboard
[00:30:09] Right. So, in this kind of a case
[00:30:14] Let me just do it here
[00:30:18] Just a moment…
[00:30:29] These are the three nodes
[00:30:31] That I will be utilizing. Now, the problem statement is right in front of you, or the example which I'm working on is right in front of you
[00:30:41] And if you take a look at it, you will understand that
[00:30:47] You know?
[00:30:48] There is something known as a user input. Now, based on my
[00:30:55] problem statement based on what I'm actually, you know, expecting from my model, you will see my user input will be nothing but the resume of the candidate
[00:31:09] Presented
[00:31:11] That as you may have
[00:31:12] candidate
[00:31:30] And, along with that, what are you going to match it with?
[00:31:34] And the job description
[00:31:37] Is that good
[00:31:38] The job description
[00:31:41] So
[00:31:42] What I will do is, I will try to pass the resume of the candidate and the job description as user input
[00:31:52] This user input will be connected
[00:31:56] to the generate node. And then the generate node will be connected to the output node. Basically, when I say generate node, in the generate node, what will I have
[00:32:08] This generate node will be responsible for generating
[00:32:14] The match sport
[00:32:19] Right.
[00:32:20] Generate the
[00:32:23] Match code
[00:32:25] And
[00:32:26] When I want to present the output, how will I present the output? Maybe I can try and present the output in the form of a web page, I can present the output in the form of a sheet, Google Sheet.
[00:32:41] So, something of that nature, I can try and do. So, as an output from my model, right, as an output from the workflow, maybe I can have a sheet, right, a sheet
[00:32:56] which act as a tracker
[00:33:00] for my HR and that sheet should have name of the candidate
[00:33:05] The skills
[00:33:07] This candidate has, right? The match score
[00:33:16] Okay.
[00:33:17] match score which has been obtained
[00:33:21] Okay. Match score which has been obtained. And final recommendation, let's say
[00:33:38] Okay.
[00:33:39] Final recommendation. So, that kind of a, you
[00:33:45] the representation is something that I can try and have. If you are not interested in having an Excel sheet, you can maybe have the same thing in the form of a web page also, right? But here, I feel especially for
[00:34:01] The HR professionals, a tracker kind of a thing, an Excel sheet if it gets automated every single time. I think that will make more sense. So that's why I'm doing it in this way, right? But there is no hard and fast rule that you will be doing it in this manner only.
[00:34:17] You can always do it in a different way, that's not a problem, right? So
[00:34:23] How am I doing it, or what kind of a, you know, situation will I have is something that you can see right in front of us
[00:34:35] Right
[00:34:36] No
[00:34:37] As I said, the resumer of the candidate and job description. So let's go back
[00:34:44] And here
[00:34:47] Here, we will see how we can
[00:34:52] Consider this
[00:34:53] So
[00:34:55] I'll go to
[00:34:56] user input
[00:34:58] I'll
[00:34:59] kind of link this user input, this one also will be linked with the generate node like this
[00:35:11] Okay.
[00:35:12] So
[00:35:14] Two inputs will be provided to the generate node
[00:35:17] Alright. And then the generate node will be connected with
[00:35:22] The output node
[00:35:24] Okay.
[00:35:25] No.
[00:35:26] What is user input
[00:35:28] User input is basically
[00:35:31] The resume of the candidate. So I can say here
[00:35:36] We attach the resume of the
[00:35:41] Candy date. Right
[00:35:45] And
[00:35:46] Just below here itself
[00:35:50] Okay
[00:35:51] Towards the bottom section, you have something known as advanced settings
[00:35:58] Click on this and instead of considering input type as any, just make it upload file. So ideally and make this input as required. So ideally what will happen
[00:36:13] At this stage, the user should be asked that yes, you are supposed to upload a file, and that file will be nothing but a, I should say
[00:36:25] You know, it will be more like PDF, right? Or a Word document, right, which we will try to upload. And what will that PDF have
[00:36:36] That PDF will have nothing but the resume of the candidate, right? So user input, this is basically where I'm taking into account
[00:36:47] The resume of the candidate
[00:36:53] Right. I'll take into account the resume of the candidate
[00:36:57] Okay.
[00:36:59] And, input type is upload file, and input is required.
[00:37:05] Okay
[00:37:07] Same way
[00:37:09] When I talk about the second one
[00:37:13] Right. When I talk about the second one here
[00:37:17] I can see.
[00:37:20] This is where
[00:37:22] We
[00:37:24] specify, or we, yeah, we specify
[00:37:30] The job
[00:37:32] description
[00:37:37] detail
[00:37:39] Okay.
[00:37:40] So, I'm doing it for a particular job description, right? So, maybe
[00:37:47] I will be utilizing my ChatGPT right I'll be utilizing my ChatGPT. And here I will just ask my ChatGPT to give me a job description for a particular job profile
[00:38:03] So I can say that
[00:38:09] I am currently
[00:38:12] Hiring for an AI architect
[00:38:18] And consultant
[00:38:21] Someone who has
[00:38:24] Let's see 10 plus
[00:38:27] years of experience
[00:38:32] AI and
[00:38:34] data domain
[00:38:37] Right
[00:38:38] And someone
[00:38:43] proficient
[00:38:50] Generative
[00:38:51] AI
[00:38:52] And agentic AI
[00:38:57] That focus
[00:39:00] On
[00:39:01] fan books
[00:39:05] Like
[00:39:06] Langra
[00:39:10] Crew AI
[00:39:13] Autogen and LangChain.
[00:39:18] Correct.
[00:39:20] consider yourself
[00:39:26] And HR professional
[00:39:32] And help me
[00:39:34] create
[00:39:36] A job description
[00:39:40] For this…
[00:39:44] Okay
[00:39:47] So I'm doing it for… I mean, something which I am coming from, background from where I am coming from. But yes, you can do it for yourself. You can
[00:39:57] Do it for your own, I can say job profile that you are currently having. Okay
[00:40:03] So, when I'm writing this prompt, by the way, this prompt is also something that I have
[00:40:08] shared in the chat box. So, when you are writing this prompt, you will see this kind of an output getting generated by the model right
[00:40:19] Now
[00:40:21] What I will do is, I will directly copy this output
[00:40:26] Which comes from GPT.
[00:40:29] And I will paste it here.
[00:40:33] Okay.
[00:40:35] I'll directly paste it here in Opal. So this is my second user input
[00:40:41] Right. This is my second user input
[00:40:48] Okay, so this is the job description I can see
[00:40:57] You see?
[00:40:59] I have attached resume
[00:41:03] I have just mentioned that we will be attaching the resume of the candidate
[00:41:09] I have mentioned that the input should be upload file kind of a thing, right? And it is a mandatory thing for us to do
[00:41:19] Job description. I have used GPT right to
[00:41:25] You know, take this job description into account
[00:41:28] Okay
[00:41:29] And
[00:41:31] I'm providing both these details to the generate node. So if you see
[00:41:37] If I double-click on the generate node, the attach resume and the job description, these two things, they are already available here.
[00:41:49] Okay, because that's how I have created the workflow from attach resume node and from job description node. I'm creating a connection for my generate node
[00:42:02] Take a look
[00:42:03] Try doing it till here
[00:42:06] I'll wait here for 2 minutes, everyone. Try doing it till here
[00:42:46] Once you are done, just type done in the chat box so that I get
[00:42:53] idea that you are ready. We'll go ahead with the next steps then
[00:42:56] Do it till here, everyone. I write down done once you are done.
[00:43:15] Yeah, Ashutosh, one question. Can we do this in mobile? Like, my office laptop is not allowing me, and somehow I don't have
[00:43:26] A personal laptop with me at this time
[00:43:30] You can try… I personally have not tried in the mobile phone. You can try, maybe, try making it in landscape mode.
[00:43:38] Maybe you will be able to do it
[00:43:40] Okay.
[00:46:05] So, Satish, in this case, for job description, you don't have to have any advanced settings. Just paste the job description
[00:46:13] Because this is something which will remain static and against this static job description, you will have different, different people write different, different resumes. So you will just match
[00:46:25] Each one of them against the same job description. So this remains stop
[00:46:31] And then it helps you in generating the score
[00:46:38] Done, everyone. I can see only one response from Chandrasekhar. What about others?
[00:46:45] Okay.
[00:46:48] Quickly let me know, so that we can go ahead.
[00:46:59] Yes, yes.
[00:47:00] Can you scroll down the job description, because there's a field of importers require the toggle button.
[00:47:06] Uh, do you need a head group?
[00:47:09] I've seen that advance sheet.
[00:47:11] Here, you don't need advanced settings, ma'am. That's what I was saying.
[00:47:16] Yeah
[00:47:17] No. Yeah, that your input is required.
[00:47:18] Do we need to…
[00:47:19] Input again. This is static, you are not taking any input from the user. This one remains as it is. User input will be your, I can say, resume, which will keep on changing. Job description is something that you provide only once, and then you
[00:47:33] Keep on, you know, experimenting with different, different resume and check which one is good to go.
[00:47:40] Uh, then what about the attached resin? It's, it's, uh…
[00:47:44] Before that setting, also, is it enabled or no?
[00:47:50] Sorry, once again
[00:47:51] That's resume input, right?
[00:47:53] Uh, then you go to the advanced setting of that,
[00:47:58] I think that brings it clean. Ah, it could request.
[00:48:01] Yeah, we are doing. Yeah, yeah, yeah.
[00:48:02] Okay. Thank you.
[00:48:06] Okay.
[00:48:08] Now, once we
[00:48:10] As to…
[00:48:11] Good, yeah, yes, yes, yes, yeah
[00:48:13] Sorry, I just connected to the system just now. I was doing it on mobile. Can you ping that in the chat window one more time, the description?
[00:48:21] I am giving you the prompt which
[00:48:25] Huh, yeah
[00:48:26] Helps me, yeah. This is something which will help you in getting the job description. Okay.
[00:48:31] Thank you. Thank you.
[00:48:36] Yeah, yeah.
[00:48:37] And, then, yeah, that was only one that I
[00:48:38] I'm good, yeah, thank you.
[00:48:39] Thank you.
[00:48:42] Yes, okay, yeah.
[00:48:43] Ashutosh, one question here. So, the job description generated for me is vastly different than what it was generated for you
[00:48:49] Would that have an impact
[00:48:52] No, I mean, that's what I said, that I'm doing it for an AI architect role, right? But you can do it for any professional
[00:48:58] I took the same prompt
[00:49:00] Same prompt
[00:49:01] Yep.
[00:49:02] Just press your output once, let me take a look
[00:49:07] I mean, even if it is taking into account the essentials, like 10 plus years of experience, you know, the roles and responsibilities LangGraph, LangChain, Autogen, Crew AI, that is also fine. Language might be different. That's okay.
[00:49:22] But overall thing, if it is still fine, still similar
[00:49:27] Let me talk it
[00:49:28] one second
[00:49:31] No, it's not allowing me to paste
[00:49:35] Okay, weird.
[00:49:40] Not allowing me to pay, but the main thing is right, like, on the top, when
[00:49:46] In the job description, your first two lines.
[00:49:55] Hmm.
[00:49:56] It says that job description AI architect consultant, rather than that, right? What I'm getting is,
[00:49:58] just directly about the role, key responsibilities, etc.
[00:50:08] Okay.
[00:50:09] That is also fine. Anyways, we are having more interested in mapping the keywords only. So if the keywords are there, that's okay.
[00:50:11] Okay, your keywords are there, line graph, etc, I can see that.
[00:50:15] 10-year experience, Langraph Crew AI Autogen, all those things are present.
[00:50:22] It is done still fine, no problem.
[00:50:24] Okay.
[00:50:27] Correct.
[00:50:28] So now, once we have the resume, I mean, we will have to upload the resume, obviously, right? And once we have the job description thing ready, let's go to the generate node. And in the generate node, what did I say?
[00:50:43] In the generate node, you will have to generate the score, right? You will have to generate the score. Now, as I said, match score is something that you require, right?But along with the match score
[00:50:59] You will have to specify a few more details, right? So, for example, in the generate node
[00:51:07] You can save let me just
[00:51:11] write a prompt here. You can say that, you are an HR recruiter
[00:51:16] Like, this workflow. I'm basically talking to this workflow, and I'm saying that
[00:51:23] You
[00:51:26] And HR recruiter
[00:51:33] Your job is to compare
[00:51:38] Okay, I can say
[00:51:40] Your job is to compare
[00:51:46] the uploaded
[00:51:49] The resume. Right
[00:51:52] this
[00:51:53] Basically
[00:51:58] Okay.
[00:51:59] This, basically. Right.
[00:52:02] With the job description
[00:52:08] Correct.
[00:52:10] Which is this
[00:52:11] Okay
[00:52:12] This is all that the generate node has to do
[00:52:18] Okay
[00:52:20] And once the generate node is able to do its part
[00:52:25] The output node here, I will click on manual layout and I will choose Save to Google Sheets
[00:52:34] Okay, I told you I need a tracker kind of a thing. I need a…
[00:52:39] Tracker sheet kind of a thing. So I will say save to Google Sheets in the output right and I will say
[00:52:52] Or maybe
[00:52:54] Let me just quickly come back. In the generate node, I'm saying you are an HR recruiter. Compare the uploaded resume with the job description, right?
[00:53:05] Here, I will just add one more line. I will say extract
[00:53:10] candidate name
[00:53:14] Okay.
[00:53:15] Skills
[00:53:18] Okay. Experience
[00:53:22] Match school
[00:53:26] And, match score between 0 to 100
[00:53:32] Right. And… recommendations
[00:53:41] Okay. And the recommendations. This is my full prompt that goes into the
[00:53:48] generate node. Okay. Just replace that metadata, guys. I have given you the prompt, just replace that metadata, which is there in the, you know, round bracket, sorry, curly bracket, and replace it with these
[00:54:04] actual input nodes, right? Or directly write it from here. It's not a very big prompt. Just quickly write it from here. You are an HR recruiter. Your job is to compare the uploaded resume with the job description
[00:54:19] Right. And
[00:54:23] extract the candidate names, skills, experience, match score, and recommendations.
[00:54:31] Okay, this is what you are doing in the generate node, because I told you that in the generate node, we will have to generate the match score, and along with that, you can always extract some more details, you can make it much more you know
[00:54:47] Structured in nature in the generate node
[00:54:50] All right
[00:54:52] And once we are ready with this as a step, what we will do next, we will go to the output node right and I will say that
[00:55:03] First of all, I will choose Save to Google Sheets here in the output node, I will say save to Google Sheets and then I will say
[00:55:14] Generate
[00:55:16] Name
[00:55:17] Skills
[00:55:23] experience
[00:55:28] Match
[00:55:30] Score
[00:55:33] And… recommendation.
[00:55:44] Okay.
[00:55:46] Generate name, skills, experience, match score, and recommendation. I mean, it has to create a Google Sheet, that's okay, that's true, right? But that Google Sheet, what all columns should that sheet have?
[00:56:00] It should have all these different sheets, right? So this is what I am explicitly specifying in my prompt itself
[00:56:11] Okay.
[00:56:14] Take a look
[00:56:16] This is my overall workflow.
[00:56:19] Attaching the resume
[00:56:21] matching it with the job description. generate the output output basically in terms of the match score. And then
[00:56:32] You know, presenting that output in the form of a Google sheet, which will have columns like
[00:56:42] skills, experience, match score, recommendation, etc. Right
[00:56:49] Take a look
[00:56:52] Let me know if you are able to build it till here.
[00:57:33] Hello. Hi, S.
[00:57:37] Yeah, yeah, hi
[00:57:38] Yeah, so, in the advanced settings, I'm not able to see this right node, actually. I see only edit each time. Or did you get one
[00:57:48] Just tap somewhere here, I mean
[00:57:51] Tap somewhere here, and then again, maybe go to the output node
[00:57:54] Okay.
[00:58:08] No, I'm not getting that.
[00:58:11] Coming, sir. Coming is working, absolutely.
[00:58:12] I'm getting this added each time.
[00:58:41] Okay, are we good? Can we go ahead?
[00:58:46] generate output and two inputs.
[00:58:58] Alright.
[00:59:01] No
[00:59:02] As a next step, we just have to
[00:59:05] Take a look at the model which it has created. We have to take a look at the application that it has created for us, right? So, on the top, you see there is this editor mode, which we were in. We were editing something, we were creating something
[00:59:20] Right, so it was in the editor mode. Now just click on the app mode
[00:59:26] Alright, click on the app mode, and it shows HR Resume Scanner Agent. This is the same name that I had given here, right on the top left corner, the name which I had provided, it is the same name that it is trying to consider
[00:59:42] Right. And
[00:59:46] This is HR resume scanner agent. The moment I click on Start
[00:59:51] It says, here we attach the resume of the candidate and it is asking me to upload your response. Upload your response in this case, it is nothing but your PDF,
[01:00:05] For a Zoom E. So I'll click on this plus option, and I'll say upload from device, right? So I have my own resume available
[01:00:15] So I'll just attach that resume
[01:00:18] Okay.
[01:00:19] And I will just click on the submit button
[01:00:26] Okay.
[01:00:27] Then it says this is where we specify the job description in detail. Let's go ahead
[01:00:34] And then it has started thinking, it is analyzing the objective, it is the generate node which has started working in the background right
[01:00:44] And, okay, it says I have received the resume, but I'm missing the job description to perform the comparison. Okay.
[01:00:53] Let me then just copy-paste it once again from GPT
[01:00:58] Right
[01:01:01] We will make some changes in the workflow, no problem.
[01:01:03] Okay, I'm directly pasting it from GPT here
[01:01:07] And,
[01:01:08] Both the things, the job description and the resume comes from the user in this case, and the generate node has started working. And finally
[01:01:20] As an output, it will give you that Excel sheet kind of a thing, which you will be able to ideate, which you will be able to keep for yourself as a tracker.
[01:01:33] Okay
[01:01:34] So, it is saying verifying practical experience, match score
[01:01:39] Right. Let's see what it gives me
[01:01:48] Generating the output, basically that Excel sheet which I was expecting
[01:01:54] Great
[01:01:57] And, yes. Now, when I click on this open in new tab option
[01:02:04] It will give me
[01:02:07] sheet like this. Okay. I will have to do some editing. Editing as in some formatting will have to be done. For example, these, you
[01:02:18] These columns, I'll have to just ensure that they are
[01:02:23] Properly arranged that much, I will have to do, right?
[01:02:30] And yeah. So for me
[01:02:33] It is giving me a match score of 85 out of 100, right? And it says that, this is
[01:02:41] As an individual, as a candidate, this is somebody who is a strong fit right exceptional alignment with core tech stack land graph, crew AI autogen interview focus verify hands-on production deployment versus training confirm totalyears of experience
[01:02:58] Okay, these are things that the HR should be asking me. Okay, so that kind of a recommendation is coming from AI, right? And skills, it has already extracted from my resume, right?
[01:03:13] And, experience is also something that it has extracted
[01:03:19] You know, from my
[01:03:23] I resume. Okay
[01:03:25] This is how
[01:03:26] automation, it works
[01:03:29] Now, what I will do is, I will again restart this okay I will again try to restart this and I will again click on
[01:03:39] start, or maybe before diet. Let's go back to this one. Okay.
[01:03:46] And,
[01:03:48] Maybe here
[01:03:52] I will make it
[01:03:55] Input type as text and this part I will directly copy paste. Okay, so
[01:04:03] I will not write it here. I will copy-paste only
[01:04:07] Okay.
[01:04:09] So, I'll just remove it from my
[01:04:12] Job description. I'll just keep it here as it is
[01:04:17] Okay.
[01:04:18] And now, if I go back to my app, and if I click on start, it says here we attach the resume of the candidate. Let me attach a different resume at this time. Okay. Just to check how well it is working
[01:04:33] I will attach another resume
[01:04:36] Okay, so I have a resume of
[01:04:40] Maybe somebody known as Vikas Kumar
[01:04:45] So I'll upload this PDF
[01:04:47] I'll click on Submit
[01:04:50] And it says this is where we specify the job description in detail, so I'll specify the job description again extracted from ChatGPT
[01:04:59] Again, click on submit. It will start thinking. It has both the details now. It has information about the job description. It knows the profile or the resume of the candidate
[01:05:14] Alright, let's see what it comes up with.
[01:05:19] So it says it is analyzing the resume and the job description.
[01:05:25] Extracting key data points
[01:05:26] Assessing skills, calculating the map score
[01:05:36] And now it has started creating the output, which is nothing but the same Excel sheet.
[01:05:43] All right
[01:05:45] And now when I click on open
[01:05:48] You see?
[01:05:50] Match score is 5 out of 100, which means again, it is able to understand that which profile is a good one, which profile is not a good one, that kind of a thing it is able to analyze and
[01:06:05] As a recommendation for this particular profile, the
[01:06:09] recommendation that comes from the model, from AI, is that this is an immediate rejection. This is somebody who is more like an entry-level in data domain. He lacks 10 plus years and GenAI expertise
[01:06:26] You can pivot him to junior data analyst or BI developer roles. But again, upskilling is required in machine learning, NLP and stuff like that.
[01:06:38] So this gets automated here. Yes, Ankit.
[01:06:41] It added the header also again, right? So, do we have to make some changes over there?
[01:06:54] Okay.
[01:06:55] We can, but I mean, these are small things we can try and edit that manually also. The reason why it is doing it is because when you take a look at… let me bring you back to the editor. The reason why it is doing is because of the output
[01:07:03] You see, every time, it will generate the name, skill, experience, math, score, and recommendation.
[01:07:09] Alright, so since the columns, they are repeated, that's why it does it. So that part, maybe we can eliminate
[01:07:15] Okay.
[01:07:18] Okay, and one more thing, actually. So, we have mentioned two different inputs
[01:07:21] Right after. So, I think the static part should actually go into the generate, right? The job description, if we actually specify it in the generate
[01:07:32] We don't have
[01:07:33] Basically, basically what happens is, you can do it, but the problem will be that in the generate node, it takes input, you see? Input variables or input containers, it takes those into consideration. So
[01:07:45] But job description is not really an input, right? It's just a static text.
[01:07:50] It is, yeah
[01:07:52] So, with the job description that we are mentioning here.
[01:07:57] Can we not actually put the static text like we have put the rest of the things, and and then generate
[01:08:01] Okay, that way. Yes, yes, yes, that is also done, yeah. So maybe what you can say is, I mean, you can say you are an HR recruiter, your job is to compare the uploaded resume, keep this attached resume part
[01:08:11] As an input, yeah.
[01:08:22] Yep.
[01:08:23] Exactly, exactly. So instead of this one, right, instead of this line, you can just have the actual job description and then finally you can say extract candidate name, skills, experience
[01:08:27] Yes.
[01:08:28] That is also done.
[01:08:32] Right. Experience this, everyone. Try doing it by yourself.
[01:08:36] I have shown you two examples, one where it is a good fit, and the other one where it is not a very good fit.
[01:08:44] Right.
[01:08:47] Try it out by yourself.
[01:08:48] Take 2 min.
[01:08:50] Upload some resumes
[01:08:52] See if it is working or not.
[01:09:00] Everyone, give it a try.
[01:09:09] Yeah. Meanwhile, I have one question. So, how is… how is it matching the score
[01:09:16] 4 is basically coming from the keywords, so it's like, when you take a look at the resume, right? When you open the resume, we always in HR and in recruitment, we talk about ATS score, isn't it
[01:09:31] So that ATS score is basically your match score, nothing else.
[01:09:36] Okay, is it defined somewhere,
[01:09:39] Not defined as such, but in the generate node, for example, if you see, it is already leveraging an agent, which will have anyways, you know, that understanding of what ATS is, how ATS is calculated, so we don't explicitly have to define it
[01:09:54] Whatever understanding it has basis that it will calculate and give it to you
[01:10:01] Okay, thank you. And one more question, when was this last class happened? Because
[01:10:07] You mentioned there was a last class which was there for the where you mentioned the basics of
[01:10:15] I don't exactly recall the date, maybe your friends can help you. But that day, I talked about N10. I think you can take a look at the recordings if you were not able to attend somehow. That… that second also, I think, was good people
[01:10:32] I think they found it beneficial. So I don't exactly recall the date. If somebody remembers the date, maybe you can put it in the chat box.
[01:10:41] I don't exactly recall
[01:10:43] Yeah.
[01:10:45] Okay.
[01:10:46] I think it was last to last weekend
[01:10:49] Last to last weekend, okay.
[01:10:52] Okay, thank you.
[01:11:02] Generate node is not working. Can you share your screen, Lokesh, once? I'll stop sharing. Can you share your screen? Let's take a look.
[01:11:16] Others also, please give it a try, see if it is helping you or not.
[01:11:28] Show me the editor part, yeah.
[01:11:32] Click on the generate node ones
[01:11:36] You are an HR recruiter? You are a recruiter? Yeah, fine, okay. Then the output node, okay? Click on the app.
[01:11:46] Okay, so…
[01:11:50] Let's start from the beginning
[01:11:52] I have the resume, correct, correct.
[01:12:00] Submit
[01:12:03] Just paste that, job description
[01:12:07] Here itself, in the type or upload response. Paste that job description that you would have extracted from GPT
[01:12:25] Submit
[01:12:30] Let's just wait for 2 minutes.
[01:13:07] Open this
[01:13:12] And a bit of editing would be required in terms of
[01:13:17] Putting it into its proper format
[01:13:20] The rest it works
[01:13:28] Okay.
[01:13:31] Yeah, yeah, sure, thanks for that.
[01:13:33] Okay, okay. And just, just make that small little change, Lokesh, that I also did. The job description, instead of, you know, waiting for it and then maybe providing it, just have it. Let me just share my screen once again
[01:13:49] Yeah, yeah, sure.
[01:13:50] I will show you
[01:13:55] So, do it this way. Job description, don't write everything here itself. Just remove that part, and you can just paste it wherever you are running this. Just paste it, or what Ankit was also suggesting, you can do it that way also
[01:14:13] Anyways, it works
[01:14:17] Okay, I hope Opal is clear to us. We are able to understand how Opel is working
[01:14:25] So
[01:14:27] In the beginning of our session today, while I was talking about the agenda and everything, I talked about
[01:14:35] the fact that we have completed… I should not use the word completed, but I have given you some understanding of NATN in the previous session, and today we have experienced opal, right?
[01:14:45] The next half an hour, 40 minutes, 50 minutes, whatever we have
[01:14:50] This is where we will talk about agentic development environment. Basically, I'll talk about cursor, right? So for that, what I would request all of you
[01:15:02] To do is just go to this website, cursor download
[01:15:10] And download the application, cursor application, right? It's not a very, I can say heavy software as such, but I would request all of you to please download this because the cloud
[01:15:25] format of cursor is paid. Right now, I'm not using anything which is paid, right? I have pasted the link in the chat box.
[01:15:35] Just use this download cursor, install it, and then we will take it ahead
[01:15:48] Okay.
[01:15:49] try doing it, I'll be back in 2 minutes. Okay, I'll just have some water. I'll be back in 2 minutes, and then we will continue from here.
[01:15:58] I think it was before that
[01:15:59] Yeah, yeah.
[01:16:00] Just one question. I think I have logged in wrong, or right, I don't know. This is, code for TV AIML
[01:16:12] Something else.
[01:16:14] Couldn't hear you once again, please, Anup.
[01:16:18] What you are,
[01:16:20] It is included in
[01:16:22] It is the part of, PG of AIML, IT, or it is something else?
[01:16:29] No, no, this is the same thing, since Dr. Durga is not available today. That's why I'm taking her, session.
[01:16:33] Okay, okay. That I received something yesterday was something else, and today it is something else.
[01:16:42] Yeah, yeah, yeah.
[01:16:44] Okay, okay, no problem. Yeah
[01:16:46] I am on that track. That's why I just wanted to test it. Thank you.
[01:16:50] No problem, no problem.
[01:16:53] Ashutar, we are expected to download cursor, is it
[01:16:58] Okay.
[01:16:59] Yes, yes, just download it on the right-hand side corner, you have this download option
[01:17:02] Just download this, install, and then we will take it ahead.
[01:17:06] Okay.
[01:17:08] Okay, I'll be back in 2 minutes, everyone
[01:20:39] Alright.
[01:20:41] I hope we are ready now, everyone. So, once you are able to
[01:20:48] Install cursor
[01:20:50] Right.
[01:20:52] Once you're able to install cursor
[01:20:55] You will be able to see on the top right corner, there is this section where you will be able to write your prompt
[01:21:07] And basis your idea that you have specified in your prompt, it will start working
[01:21:14] Okay
[01:21:16] So again, how that really works is something that we will take a look at. All right. But again, before doing everything, what you should do is
[01:21:27] You should go to your
[01:21:29] I can say desktop. I hope my desktop is visible to all of you. You can go to desktop
[01:21:36] And just create a new folder
[01:21:40] Okay.
[01:21:41] You can go to your desktop, and just create a new folder, right? And you can give it a name, maybe you can call it cursor demo
[01:21:55] Okay.
[01:21:56] You can call it cursor demo
[01:22:00] And
[01:22:02] This particular cursor demo is where you will be able to
[01:22:08] Perform and ideate your things
[01:22:12] So for doing that, what you have to do is
[01:22:14] Just open your cursor application, right? And here
[01:22:19] It is saying open project clone repo connect via Ssh right?
[01:22:28] We will click on open project. Right. And I will just open that particular folder called cursor demo
[01:22:41] Okay.
[01:22:42] So I'm opening that cursor demo
[01:22:46] inside
[01:22:47] my cursor application, you see, this is how your
[01:22:52] interface that should look like. On the right-hand side corner, you should have this new agent
[01:23:00] Okay, you should have this new agent. This is where you will write your prompt, and on the left-hand side, you should see your cursor demo
[01:23:11] You know, folder.
[01:23:13] Right, so we are opening the cursor demo inside
[01:23:18] cursor application, okay? That new folder which we have created, we are opening that folder inside cursor application
[01:23:30] Alright. And those of you who are already familiar with Visual Studio Code, you'll find that cursor looks almost similar to Visual Studio code
[01:23:45] Right
[01:23:46] So those of you who have some idea of VS Code, you'll find that. But anyways
[01:23:51] This is what needs to be done. So install cursor, right? So what are the things that needs to be done? Download and install cursor
[01:24:06] Second thing that you have to do is
[01:24:10] create
[01:24:11] new folder
[01:24:15] And
[01:24:17] Give it a name
[01:24:19] Right.
[01:24:20] Third
[01:24:22] go to cursor
[01:24:26] application
[01:24:27] Right. And open that
[01:24:32] folder
[01:24:34] created
[01:24:38] created in step 2.
[01:24:41] Okay. If you're not directly able to open that particular folder, on the top left-hand side screen, you have this option called File
[01:24:52] Click on file, click on open folder, and from here also you can open the same folder which we have just now created called Cursor demo
[01:25:05] Quickly do this
[01:25:12] Hey, everyone.
[01:25:15] Please quickly do this.
[01:25:43] It is showing new chat, not new agent. Can you share your screen once, Rutika?
[01:25:52] I'll stop sharing.
[01:25:53] Yes.
[01:26:10] Okay.
[01:26:11] Yeah, it's showing new chat.
[01:26:13] Just trying new chat
[01:26:20] Can you ask continue to log in.
[01:26:22] Just click on that login option, see if that is needed
[01:26:28] Click on login
[01:26:32] There itself, we have that option.
[01:26:48] This is exciting.
[01:26:51] Is it not working? That option?
[01:26:52] near sign in, I think.
[01:26:56] Okay
[01:27:00] You might have to log in, that's my… that might be a problem. Just try once again, maybe? Or maybe just close this tab, open it once again, try logging in once.
[01:27:22] The installation was done properly, right?
[01:27:23] Yes, same link.
[01:27:28] Okay.
[01:27:36] Maybe something to do with login. We can… we can check, just try and close this tab and maybe open it once again
[01:27:46] Once you log in, I think it will work, no problem.
[01:27:47] Okay, I'll try.
[01:27:49] Okay, meanwhile, I'll just explain the next step. You can take a look. Let me share my screen once again
[01:27:56] So
[01:27:58] Yeah.
[01:27:59] This is what
[01:28:01] It looks like, right now. Okay. Now, idea is very simple. You have a plan in mind. You want to maybe create a website, you want to create a dashboard. Great, whatever
[01:28:17] That particular thing, that particular thought of yours, that particular, I can say
[01:28:22] You know, intent of yours is something that you can put here
[01:28:27] Within new agent, right? And the moment you are able to do that, okay
[01:28:34] the model, or, you know, the ADE, rather, the agentic Development Environment, will take care of the next steps. So we'll take a look at what exactly are we planning to build and how are we trying to do that
[01:28:49] I'll show it to you in a while, but that's basically how the process it will look like. Yes?
[01:28:58] Yeah, sorry, while downloading the browser app, my screen went blank, I don't know. Not only I got set up ready. After installation, what are the steps completed
[01:29:09] I have pasted in the chat box, 3 steps. After you are done with
[01:29:14] installing cursor. Create a new folder, give it a name, and go to Cursor application and open that folder created in Step 2.
[01:29:24] Okay, thank you. Let me go
[01:29:26] It's dead in the checkbooks
[01:29:31] No.
[01:29:32] If I come back and go to my Microsoft whiteboard once again.
[01:29:39] This is where I'll make you understand what exactly are we trying to build
[01:29:45] And then, I'll show you how we are building this, right?
[01:29:51] So…
[01:29:52] What is the objective? How are we doing it?
[01:29:54] Let's say that
[01:29:57] I mean, again, this is something which is cross-functional in nature, right? It is something where you will notice that
[01:30:06] You know
[01:30:08] Just a moment
[01:30:14] Yeah, so this is something which is cross-functional in nature. You can be presenting this as an output to
[01:30:24] You know, anybody, for that matter. Let's say you are creating a dashboard. Okay, CEO level dashboard so that they can make some analysis on top of it.
[01:30:40] Something of that nature. Okay.
[01:30:42] So
[01:30:44] How do you do it? And how do you build this as a model
[01:30:51] Let's try and understand. Same way, let's say, you know, you are
[01:30:57] willing to create a website altogether. So how do you do all those things is something that we need to take a look at
[01:31:07] Okay
[01:31:08] So, first of all, it becomes important for us to understand that what kind of
[01:31:17] Let's say
[01:31:19] You know, thing we are willing to create. What is it that we are willing to create? That understanding should be there, right? And the moment we have that understanding, we can get into the
[01:31:34] Next steps. So let's say with cursor, what I'm willing to do is I am willing to create
[01:31:42] Right. Now, the moment I say dashboard
[01:31:47] Okay, the moment I say dashboard, couple of things they should come to your mind. For example
[01:31:55] When I say a dashboard, you should try and think that
[01:32:01] This dashboard should
[01:32:03] be created for which industry, let's assume, right? So, let's say my end goal is that we want to
[01:32:14] Build
[01:32:16] An executive summary dashboard.
[01:32:19] Right.
[01:32:21] executive
[01:32:24] Summary
[01:32:30] dashboard
[01:32:35] Right.
[01:32:37] Now, when you are building an executive summary dashboard, right
[01:32:41] Few questions you should ask yourself. You should ask for which
[01:32:46] industry
[01:32:48] Are you doing it for? Because different, different industries, they will have different KPIs, right
[01:32:58] Who is your intended audience? Who will take a look at this dashboard? So, who is your audience?
[01:33:09] Are you doing it for external client? Are you doing it for stakeholders, or are you doing it for your internal leadership?
[01:33:21] that understanding should be there, because accordingly, you will try to process or mask the sensitive data, the private data, right? So who is your intended audience? Which industry are you referring to
[01:33:35] Okay, what is your objective
[01:33:38] You are creating a dashboard, but what is the purpose of creating this dashboard? Do we have to get a quick glance of the, you know, developments that are happening, or do we have to make some decisions
[01:33:54] Out of it, correct. What is your overall objective? That kind of a thing.
[01:33:59] It's something that you should ask yourself and then
[01:34:05] Fourth important thing is the components. As I said, each and every industry, they will have different components. They will have different KPIs, right? So based on your industry, what all components
[01:34:21] Would you like us to represent in the summary dashboard?
[01:34:29] Right.
[01:34:30] KPI
[01:34:32] Basically
[01:34:34] Okay
[01:34:35] This is something that you will have to think for yourself
[01:34:41] Okay, this is something that you should
[01:34:43] think for yourself. Now, the moment you have answers to these kind of questions, you can write a prompt okay and that prompt will be able to
[01:34:54] help you in getting the things done
[01:34:59] Okay. So, what I will do is, I will write a prompt here itself for all of these, right? You can take a look at it, right? And you can understand how do we go about it, right?
[01:35:13] So
[01:35:14] Objective is simple. You are trying to understand the fact that how exactly are we planning to build the dashboard? What are the granularities that we
[01:35:29] Willing to take into consideration, right? So I can say that
[01:35:35] It is supposed to create
[01:35:39] Create
[01:35:41] Interactive
[01:35:44] dashboard, right?
[01:35:47] It is supposed to cursor as an Ade is supposed to create an interactive dashboard right?
[01:35:57] Or maybe I can say
[01:36:01] A boardroom ready
[01:36:02] Okay, because this is something that I would like to present to my senior leadership. So I'll say, create
[01:36:11] Board room
[01:36:14] Ready.
[01:36:16] interactive
[01:36:22] executive
[01:36:25] dashboard
[01:36:27] Visual. Using
[01:36:30] Python
[01:36:32] And streamlit
[01:36:35] Okay. Python being your backend programming language, Streamlit being a framework which will help you in
[01:36:44] Taking a look at the user interface, which will be able to help you, you know, experience the
[01:36:52] dashboard in real time, the interactive dashboard in real time. Okay. Now
[01:37:01] I'll give some context. I'll say that the industry is
[01:37:06] Telecom.
[01:37:09] Okay, or maybe I will simply say that,
[01:37:15] This is to be created
[01:37:21] Telecom
[01:37:23] company
[01:37:26] Okay.
[01:37:28] This is to be created
[01:37:33] a telecom company, right?
[01:37:36] And has to be presented
[01:37:41] In front
[01:37:44] CEO
[01:37:46] And
[01:37:48] rest of the
[01:37:52] leadership
[01:37:54] team
[01:37:56] Okay.
[01:37:58] For
[01:38:01] Quick
[01:38:02] decision
[01:38:04] making
[01:38:07] purpose
[01:38:09] Okay.
[01:38:10] For quick decision making
[01:38:15] Right.
[01:38:16] This is something which needs to be
[01:38:19] Done.
[01:38:22] Right.
[01:38:23] No
[01:38:24] I'll get into the granularities. I will say that
[01:38:28] The objective, the audience, and the industry is mentioned. Now let's talk about the components. So I can say in the dashboard
[01:38:39] Okay. Include the following
[01:38:46] Okay. Include the
[01:38:49] revenue
[01:38:53] Brent?
[01:38:55] Of the last 12
[01:38:59] Months
[01:39:02] It should have customer
[01:39:06] churn rate
[01:39:09] It should have
[01:39:11] customer acquisition
[01:39:13] Acquisition ACQU
[01:39:16] I-S-I-T-I-N, acquisition cost. This is called
[01:39:23] CSC, in telecom domain, this is something which is quite an important factor, especially for leadership teams, right? And similarly, another very important factor is ARPU. ARPU stands for average revenue per user
[01:39:38] So, average revenue
[01:39:42] A user, ARPU
[01:39:47] Right? So, in the dashboard, include the following, these four things they should be incorporated right
[01:39:54] Include
[01:39:57] trend
[01:39:58] Addles
[01:40:01] Up and down
[01:40:04] arrows, basically.
[01:40:07] Right.
[01:40:14] And
[01:40:16] include
[01:40:18] Risk indicators
[01:40:22] Maybe you can say that it should highlight
[01:40:28] John
[01:40:29] If it is high, because again, for telecom as a domain, you would notice that
[01:40:36] churn becomes a very important factor. If I am a telecom provider, and if the customers, they are leaving my services, I should understand why are they leaving my services. Is it because of instability in
[01:40:53] You know, features, or in the, you know
[01:40:56] Subscription that I am providing, or is it because of the cost? What is it, actually, that is, you know, promoting my customers
[01:41:11] leave my services, and maybe opt for services from somebody else. So that's why highlighting the churn is important here, right
[01:41:22] And then I can give some designing instruction. I can say keep the
[01:41:30] Overall
[01:41:33] Background
[01:41:35] As
[01:41:36] And use mustard
[01:41:41] yellow
[01:41:43] And
[01:41:44] I should say
[01:41:47] Mustard yellow, and
[01:41:49] Lemon
[01:41:52] green color
[01:41:55] Okay.
[01:41:56] to
[01:41:59] Showcase
[01:42:00] The data
[01:42:02] And
[01:42:04] Use red color
[01:42:06] for
[01:42:08] Highlighting risk.
[01:42:13] Again, this is up to me. This is completely based on my preferences. You can always change this. You can always, you know, use your own colors, right? And I can say
[01:42:25] Keep
[01:42:27] Upgrade
[01:42:28] layout. Okay. With clear
[01:42:36] Okay, it should have grid layout
[01:42:40] And
[01:42:41] It should have clear sections very nicely maintained.
[01:42:44] Okay.
[01:42:51] Making it
[01:42:54] Instantly
[01:42:57] To read about and
[01:43:00] nature
[01:43:02] Correct. And, along with these details, I will say
[01:43:10] Add
[01:43:13] Dynamic
[01:43:17] Wherever
[01:43:19] required
[01:43:22] Okay.
[01:43:27] This is my overall prompt
[01:43:29] I'll paste the prompt in the chat box
[01:43:34] Okay. And the exact same prompt, I'm going to paste in cursor also
[01:43:44] Right.
[01:43:46] And I'll just hit enter
[01:43:51] It will start
[01:43:52] Writing the code for me, it will start building the entire interactive
[01:43:59] dashboard for me, right?
[01:44:02] And it might ask you for a few confirmations, right, in between. Just allow them, right? And you will see a working prototype ready in front of you
[01:44:31] Let's see how the output looks like. Let it work
[01:44:53] The whole Python file is something that it is creating for us
[01:45:07] You can make dashboards, you can make websites, anything.
[01:45:14] I'll give you one example for a website also.
[01:45:18] But yeah, let's first finish this. Let's take a look at the output
[01:45:28] Hello?
[01:45:30] What about? Like, is it the local LLM, or
[01:45:37] Did I set up anything else
[01:45:39] Yeah, so what happens is, when you are talking about cursor free version that we are using right now, here we don't get an idea of which LLM it is using. It simply says auto
[01:45:51] Right? But if you are, let's say, using the paid subscription of cursor, you can actually choose from different, different models. And since Cursor is a third-party application, you can play around with
[01:46:04] Any of the LLM, you can use Claude also you can use, you know, GPT models also Gemini also everything.
[01:46:13] Great.
[01:46:14] What about that all am,
[01:46:18] Well done.
[01:46:19] Polama is different. Polama is for local diploma. That is different. You can leverage LLMs here. Polama is not an LLM.
[01:46:27] Okay, then, for me, showing setup Cloud Agent 3, that's why I'm confused in the application
[01:46:36] Okay, no problem.
[01:46:39] So, for example, once this gets done, once, you know, it is ready, and if I just expand this, it says within cursor demo, I have something known as app.py, right? Now this app.py is
[01:46:55] Basically, my entire Python code, which it has written for me.
[01:47:01] I have not written anything. It is the AI agent which has done this for me, right? Now what I need to do is I just need to run this and see how it is working. Okay, so for that, what I will do is I'll go to terminal
[01:47:18] I'll click on New Terminal and here I'll just specify a command
[01:47:25] I'll say Streamlit
[01:47:30] app.py
[01:47:33] So, in your local, I can say browser, it will automatically open the dashboard
[01:47:45] And this is how the executive dashboard
[01:47:49] It looks like
[01:47:51] completely automated in nature
[01:47:54] Alright. And… 100% interactive.
[01:47:59] Alright. You can take a look at the churn rate, the customer acquisition cost, the ARPU, revenue, everything, right? And it has taken into consideration couple of
[01:48:16] Regions like the APAC
[01:48:20] Is your Pacific region, right? The Europe, Middle East region, North Americas, all these different, different regions it has taken into account. And let's say, if I'm only interested in the APAC and
[01:48:34] You know, Europe and Middle East region, let's say, if I'm not interested in the LAM and in the North America region, you will see that the
[01:48:43] Dynamic filters, they are designed in such a way that the values, they are changing. The visuals, everything, they are changing. Customer, also, I have segmented them into enterprise, consumer, etc. So let's say if I'm removing SMB right now
[01:48:59] Correct. And maybe I'm removing the Enterprise customer also, if I'm just talking about people like you and me, right?
[01:49:09] If I'm talking about
[01:49:11] the customers only. This is how the model looks like.
[01:49:17] Okay.
[01:49:20] Completely interactive in nature.
[01:49:22] And
[01:49:24] I mean, the backend gets created using Python, you have the code right in front of you, and you can just see how the model, it works
[01:49:33] Okay.
[01:49:35] This is PR code, streamlit Run
[01:49:39] app.py. Let me paste it in the chat box also.
[01:49:46] Springlet
[01:49:49] Run
[01:50:00] Try doing it
[01:50:58] So, it is in your local, right? Right now, this application is running in your local.
[01:51:04] Yes, yes, local Austria
[01:51:06] So, if you want to host this one in AWS, then you need to copy this PoE and host it, right?
[01:51:12] Yeah, on AWS, or otherwise here also, if you see on the right-hand side corner, you have this option of deploy, right?
[01:51:22] So, you can deploy it as a Streamlit application itself, right? Otherwise, what you are also suggesting can be done. That's possible.
[01:51:33] Thank you.
[01:51:39] So, above, in your case, if it says the term streamlet is not recognized, just copy the same message that you have pasted in the chat box, right? Paste it back here again inside chat interface
[01:51:55] And, cursor will take care of it. It might not have installed it, right? So, that's why it is throwing you that error. Otherwise.
[01:52:05] Come and rest is fine. Just paste that error in the interface, in the chat interface there. It will help you.
[01:52:15] Once it has done, just go back to your app.py, go to terminal, open a new terminal, write down the same command, streamlitrun app.py, it will help.
[01:55:35] Is it working, everyone? Are you able to see some output?
[01:55:39] From interactive dashboard
[01:56:09] If you're getting some error, just put it back into the same chat interface here
[01:56:17] Everyone
[01:56:18] It will rework on the call
[01:56:28] For me, it's trying to install the packages
[01:56:30] But,
[01:56:31] Yeah, yeah, just allow them
[01:56:33] Yeah, I gave it, but it still says waiting
[01:56:37] Looks like it takes time
[01:56:50] Okay.
[01:56:51] If it is taking a lot of time, then maybe, I mean, you have the prompt anyways with you. Just open a new folder, right, and just repeat the steps, because with cursor free version, it
[01:56:53] slightly lags at times.
[01:56:55] Okay.
[01:56:56] So if it is taking a lot of time, then just repeat the process, have a new folder, open that, paste the prompt
[01:57:03] And it will work.
[01:57:11] Ask those one question. Like, this dashboard has been created with some random data, right? For those revenue trend, customer channel rate, and everything.
[01:57:21] So, will it be possible, like, can we provide some dynamic, means real-time data to that, and dashboard will be created according to that? Is it possible?
[01:57:32] Yeah, so there are two answers to this. First is, with cursor pre-version, if you see, you don't get the flexibility of uploading anything, right? You can maybe upload an image, but you cannot upload a CSV file
[01:57:47] Okay, okay.
[01:57:48] That is the limitation of cursor-free, ADE. But, we also have a hack here. What we usually do is
[01:57:58] I mean, since it is completely based on our prompt, it is completely based on our intent, right? When I'm saying that it should have a revenue, you know, trend, it should have these dynamic filters, everything, what I will do is
[01:58:13] It should… I'll explicitly specify in my prompt itself that the dashboard, it should have
[01:58:19] Upload option, right, where the user is able to upload a CSV file or an Excel sheet, specify the format that you are interested in, where the user is able to upload the file right in real time and basis that if there
[01:58:37] Changes required in the dashboard. It can happen on the fly. So that kind of an upload option will also be there in the dashboard itself. The moment you upload your data, you will see that the value is there start changing
[01:58:50] Okay, okay. Thank you.
[01:58:52] So directly, it won't be possible, but yes, indirectly, you can do it by utilizing this hack
[01:59:00] Okay.
[01:59:06] Yeah, yeah
[01:59:07] Thank you.
[01:59:13] Ashutosh, one question. Can we use the cursor for generating SQL queries
[01:59:19] SQL queries also… I mean, you can do an end-to-end prototyping. So, here, it's Python which I specified, right, in the prompt, which is why it is giving me. py file, right? If I, let's say, want to do it with, let's say, Tailwind CS
[01:59:39] I want to do it with JavaScript, I want to do it with React. No problem. Accordingly, the code will get revamped
[01:59:44] So, events, if you want something, you can just specify that in your prompt and see how it is working for you
[01:59:52] Okay, and can we use, like, the code generation part for… from Akasha, the way we use Copilot
[02:00:05] Yeah
[02:00:06] Why not? I mean, you want to use this code file somewhere else, that's what you're saying, right? Yeah, yeah, yeah.
[02:00:08] Definitely, can be done
[02:00:12] See, this is something which is for prototyping purposes, for faster prototyping purposes, so that you are not spending a lot of time in writing the code and, you know, all such things. But when it comes to your own learning, right, for example.
[02:00:29] When I meet you guys regularly in the Agentic AI module, right, I'm supposed to take that. So when I meet you there, I'll be writing each and every code line by line. The reason why I will do it, I mean, now you understand that I know how to get it done using these AI tools. I am myself showing it to you
[02:00:48] But in agentic AI module, I will never use it. And for that matter, I will not even encourage you guys to do it. Simple reason is because there you should have a different objective altogether that yes, I want to understand each and every line of the code. And once my learning is done, once I know the
[02:01:07] Weigh how, let's say, LangChain, LangGraph, whatever is working. I can always have these tools for me who can, you know, do this thing. But let's say
[02:01:16] You don't know how these frameworks, they work. For example, there might be people in the room who might not know that how Streamlit works. Now, if I directly show you the Streamlit part of the code, you might not be able to make sense out of it. That should not happen, isn't it?
[02:01:32] That should not happen. So, from learning perspective, you should always try and do it manually, the way how we will do it in the sessions. But again, when it comes to prototyping and maybe showcasing something, like, how much time did it take
[02:01:47] I think 10-15 minutes, and I was ready with the dashboard. So let's say when you want to do faster prototyping, just use these tools, don't keep idle, and don't just write things manually then. Be a bit smart. But yes, at the same time, you should
[02:02:03] Knowing how each and every line works.
[02:02:07] Yeah.
[02:02:08] Yeah, makes sense.
[02:02:14] Okay, so I think for most of us, it is working. Just in case if you feel it's not working for you, again, as I said, with cursor free version, it is slightly lagging at times, that's basically a limitation of the free version, but you can always
[02:02:30] Repeat the steps. You have the prompt with you, right? You can create a new folder, open that folder, put that prompt, and it will work for you, alright
[02:02:39] 95% of the times it will work directly. But again, for 2, 3, 5% of the times, if it's not working, just repeat the steps and it will help you, right?
[02:02:50] So with this, at least you understand that how the paradigm shift
[02:02:55] is looking like right now. How we can create interactive dashboards using these models, and then, obviously, how things like Opal, they can help us, right? So we have tried and tested
[02:03:11] ourselves with both Opel N8 and ADE, like, cursor today, right?
[02:03:17] Any questions if we have? I think we can discuss, and then, since it's almost time, we will conclude for the day.
[02:03:32] Any questions, anyone?
[02:03:34] You were telling about some website designing and all that will be laptop maybe, due to this time constant.
[02:03:41] Website, I'll just tell you how you go about it, right? And then maybe, because at the end of the day, it is just about your prompt, right? So, what we can try and do is we can basically, first of all, figure out that
[02:03:57] Why are we creating a website? For what purpose are we creating a website? That idea should be there in our mind, right? For example.
[02:04:09] Let's say you are creating a business strategy simulator, right? Now, this is something which consultants like me, they use a lot. When I say a lot, I personally use these kind of things every week because let's say whenever I'm trying to interact with my clients
[02:04:27] They need a specialized roadmap. They need a specific roadmap for their own industry, for their own problems, right? And the template that I, as a consultant, use might work for one particular industry, might not work for any other
[02:04:44] Industry, right? So, what business strategy a particular company or a particular client should use, and what it should not use, should bevery clear, right? So, for me, as an individual, I can create a business strategy simulator
[02:05:01] That will be more like a website, where I will ask the client himself or herself to provide the data, data like which industry is is the client coming from
[02:05:15] company size, annual revenue, right, market region, etc, and basis that it can do the needful for me
[02:05:23] A couple of use cases they can be developed and accordingly
[02:05:33] platform cursor. It can generate the output for you
[02:05:41] Yeah.
[02:05:42] Okay
[02:05:43] So, let me… okay…
[02:05:47] Any links, links, so I don't have
[02:05:54] Let me just show you, it won't take me much time.
[02:05:59] It's almost 12 o'clock, but let me just show you.
[02:06:03] Let me create a new folder on the desktop here, and let me call it cursor demo
[02:06:15] Right.
[02:06:16] And, I'll open this cursor demo 2 inside cursor now. So I'll click on File, I'll click on Open Folder
[02:06:26] Right. And I'll open this cursor demo to new folder I will open
[02:06:33] Right.
[02:06:34] And the moment I open the new folder here, I can maybe same thing, write a prompt, and that prompt will be helpful in, you know, generating the
[02:06:48] Website maybe this time. So if I want to create a website, how do I do that? Let me just
[02:06:54] Show it to you. So, my objective is very clear. I want to create a business strategy simulator kind of a thing right? So I'm just writing the prompt here first
[02:07:08] And then I will paste it into cursor directly
[02:07:12] Okay, so
[02:07:15] And this will make you understand how websites can also be created
[02:07:21] Okay, so what I will do is, I'll say…
[02:07:27] Build a clean
[02:07:30] interactive, interactive
[02:07:33] consulting
[02:07:36] Strategy, strategy
[02:07:40] simulator.
[02:07:43] website
[02:07:44] Right. This is the, task that it has to perform. This is the instruction that I'm giving this model
[02:07:55] Alright. And then I am saying that, this website for, let's say, management
[02:08:04] Consultants
[02:08:06] Right.
[02:08:09] Okay.
[02:08:10] For
[02:08:12] simulating. Simulating
[02:08:16] Maybe, what? For simulating business strategies
[02:08:21] Business Strategies
[02:08:27] Client
[02:08:28] Okay.
[02:08:30] For clients
[02:08:34] the website
[02:08:40] a section
[02:08:43] For user input
[02:08:46] Basically.
[02:08:48] Industry
[02:08:52] company size
[02:08:54] Okay.
[02:08:55] And here, who is your user? User is basically your
[02:08:58] I can say client only
[02:09:01] Great. So, client will specify his or industry, company size will specify annual revenue
[02:09:13] We'll also specify
[02:09:16] Business problem
[02:09:19] We'll also specify market
[02:09:23] Region? Okay
[02:09:27] Business problem, market reason, and maybe business problem, I will make it as a drop-down, so I will give some usual business problem that I usually see with my clients, right? And it can just choose
[02:09:41] from one of those business problems.
[02:09:44] I can say that after
[02:09:47] Clicking
[02:09:50] An option called, generate strategy
[02:09:56] Okay
[02:09:58] The website…
[02:10:00] Should show
[02:10:03] The following. It should show the problem diagnosis
[02:10:10] Okay.
[02:10:11] Petrol the problem
[02:10:15] Diagnosis… wait a minute…
[02:10:23] Diagnosis
[02:10:26] It should also
[02:10:28] Sure. Okay.
[02:10:30] In the generate strategy, it should show strategic options
[02:10:36] Strategic
[02:10:39] Options
[02:10:40] It should also show
[02:10:43] short-term actions
[02:10:46] It should show long-term
[02:10:50] Transformation ideas
[02:10:55] Right. These are the things that it should try and showcase
[02:11:05] Let me… make it slightly
[02:11:19] So
[02:11:21] It should try and, show long-term transformation ideas
[02:11:26] And,
[02:11:28] Key the risks
[02:11:30] Okay
[02:11:32] The key risks they should be highlighted. Now, the way how I was creating a dashboard, right, I was specifying the format and everything, right? This time what I will do is I will say key
[02:11:46] the UI
[02:11:49] Modern and consulting style
[02:11:54] Consulting style. By consulting style, I mean it should not use a lot of bright colors, right? It should be minimalistic in nature, right?
[02:12:08] White
[02:12:11] And
[02:12:13] teal green
[02:12:15] team
[02:12:17] Okay.
[02:12:22] Use the following
[02:12:26] technology, or tech stack
[02:12:29] To achieve
[02:12:34] Yes. So, since it is a website, I will use HTML, right? I will use Tailwind
[02:12:43] CSS. I will use JavaScript
[02:12:48] JavaScript. Right. Somebody, if you want to use any different technology stack, you can please do that. I'm just giving you an idea how we can create it here
[02:13:03] Right. And then I will say that,
[02:13:08] as an output
[02:13:12] I am expecting
[02:13:15] A complete
[02:13:19] working web application
[02:13:23] Right.
[02:13:25] Completely working web application
[02:13:27] with separate, separate
[02:13:32] Section
[02:13:37] Head up
[02:13:40] Input
[02:13:42] Well
[02:13:45] dessert dashboard
[02:13:50] And some footer
[02:13:53] Include a sample mock strategy
[02:13:59] output. Okay. So, if somebody is very new to this kind of a thing, doesn't really know how to use this, how to leverage this right
[02:14:10] For those people, I will have a sample mock, right? And that's it.
[02:14:17] This is my overall prompt
[02:14:19] I'm putting my prompt in the chat box also
[02:14:24] Right. And the same prompt, I'll go back to cursor, and I'll paste it here
[02:14:30] And when I hit enter
[02:14:33] It will start planning the next moves, and maybe in the next 5 minutes, it will be ready with the HTML file
[02:14:42] Which I will just open and show you the output from
[02:14:50] Let's just wait for a while
[02:14:59] Ashutosh, one question.
[02:15:02] When we… let's suppose it generates the output, and when we want to little bit update that, so do we have option in the cursor to
[02:15:15] Yeah, yeah.
[02:15:16] I mean, update the output, like, the way we have initiated follow-up questions.
[02:15:18] Correct, correct. So what happens is, in, Karser, we have inline updation, right? So, for example, if there is a code, and if you want to change that code, right, you can always do that
[02:15:33] I mean, again, very simple. For example, let's say if I just open this index.html, right? You can see the entire code file. It is ready right in front of you. Now, if you want to change anything, okay, let me just show you. For example.
[02:15:51] Or maybe this JavaScript file
[02:15:56] app.js. Alright. Whatever, alright. If I have to give some additional thing, I have to specify the risk something something right what I'll directly do is I'll
[02:16:13] I'll just maybe copy this part, right?
[02:16:17] I'll copy this part, and there is this option called Quick Edit, right? The moment I click on Quick Edit, I can write a prompt here
[02:16:27] And I can say that, again, what is to be done, what is not supposed to be done, if it has done something which you were not expecting, just write it here itself on the fly, and it will modify that particular highlighted region
[02:16:42] Okay.
[02:16:44] So, that's how it works. Let me just bring us back to the
[02:16:52] folder
[02:16:53] Cursor demo tool. Here you see your,
[02:16:58] HTML file, your JavaScript file, everything is ready. If I just click on this index.html
[02:17:06] This is basically the simulator that it has created for me. So it says that it is an interactive playbook which
[02:17:14] Captures engagement context, generate a structured strategy, etc. So for example, I will specify my industry as
[02:17:24] FMCG
[02:17:25] Okay. Company size, let's say I have 1000 to 499 employees
[02:17:33] Our revenue is let's say 50 million business problem is maybe digital and technology modernization. Market region is
[02:17:44] Asia Pacific
[02:17:46] And I'll just click on Generate Strategy
[02:17:50] So, for industry FMCG business problem this much, market region this, it is doing a problem diagnosis. It is giving me strategic options 1, 2, 3, short-term action, 90-day action plan basically, and then long-term transformation ideas
[02:18:06] And what can be the risks associated, right? Let me change it. Let's say industry this time is telecom
[02:18:17] I have
[02:18:20] 250 to 1,000 employees. Annual revenue is, let's say, this much
[02:18:26] Business problem is…
[02:18:29] Maybe…
[02:18:32] Accelerating revenue growth
[02:18:35] Market region is Middle East and Africa. Generate strategy
[02:18:40] You see? In an interactive manner, you can just keep on playing around with this.
[02:18:48] And for each and every
[02:18:51] unique client. It will give you strategic recommendations, which people like me can always use and can propose to our management consultants.
[02:19:04] This is basically what management consultants, they usually do
[02:19:11] Yeah.
[02:19:12] This is a working website for you.
[02:19:17] Messages emulator.
[02:19:19] Thank you.
[02:19:25] I think this gives us a fair idea of how we are utilizing
[02:19:37] I've got Flutter as an application. I have shared the feedback also with all of you. Let's take a moment and please share your feedback. Let us know if it is
[02:19:48] helping you, if it was, you know, able to add some value to your
[02:19:55] Alright.
[02:19:57] Yeah, this is basically what we had to cover today.
[02:20:01] As I mentioned last time also, I'll be meeting you guys very soon, once we are into the agentic AI module. I'll be responsible for handling it end to end. So there we'll talk about the technical frameworks, all those things in detail
[02:20:17] But yeah, even here, once you are, let's say, slightly comfortable with the no-code frameworks, once you understand and start appreciating the capabilities of generative AI and agentic AI, when we meet in the agentic AI technical module, there we'll try and see how this actually
[02:20:39] So, that's what we have.
[02:20:43] Anything else, if we have to discuss, maybe please
[02:20:49] Maybe we can call it a day.
[02:20:50] Hello?
[02:20:53] Yes, it is, yeah.
[02:20:54] Yeah, uh, sorry, uh, just it will take up one minute only. I was able to ask that, uh…
[02:20:58] Uh, can we connect, uh, just like some options are there? Can we connect with the MySQL database?
[02:21:04] But it's…
[02:21:10] Okay.
[02:21:11] I have a couple of connectors now that is, again, we'll have to check what all connectors are available based is that you can, for example you can connect it to MySQL database. You can even connect it to a couple of plugins that are, let's say, allowed. So that is also possible.
[02:21:21] Okay.
[02:21:23] Oh, I don't see feedback, can you share it with me?
[02:21:28] It's a tool, basically, Lakshmi
[02:21:32] And see if you are able to see a couple of
[02:21:35] You have already responded.
[02:21:40] It's not visible to me, Sama.
[02:21:45] Okay, no problem.
[02:21:47] 30 people, they have been able to
[02:21:51] share their feedback. But anyways, no problem. I hope you were able to follow.
[02:21:57] Yeah.
[02:22:01] Alright, so in that case, we'll then keep it till here. Thank you, everyone, for joining.
[02:22:08] We'll, I mean, you'll be able to continue with your regular sessions with Dr. Durga from the next week. And once we are in agentic AI there I'll certainly meet you once again.
[02:22:24] Thank you.
[02:22:25] Okay, thank you. Bye. Take care.