Transcript for transcript_1dTwghJ0-CnMliqHXLRdq2KsrMIauIRjp_1_2700_9524509ffbbc_1.txt

transcript_1dTwghJ0-CnMliqHXLRdq2KsrMIauIRjp_1_2700_9524509ffbbc_1.txt

[00:16:31]  Hello.
[00:16:33]  Hello, sir.
[00:16:35]  Hi, Tripoon.
[00:16:36]  How are you?
[00:16:37]  I'm good, sir.
[00:16:38]  How are you, sir?
[00:16:39]  All good, all good.
[00:16:41]  Where are you right now?
[00:16:42]  Which city?
[00:16:43]  I'm at Hyderabad, sir.
[00:16:45]  Okay, Hyderabad.
[00:16:46]  Okay.
[00:16:47]  Great.
[00:16:48]  So, what's up?
[00:16:50]  What's up happening nowadays?
[00:16:53]  What are you doing?
[00:16:54]  Your college is over?
[00:16:56]  Like, when is it over?
[00:16:58]  Yes, sir.
[00:16:59]  I have passed out in 2025, sir.
[00:17:01]  25, right?
[00:17:02]  Yes, sir.
[00:17:03]  And you're...
[00:17:07]  Just give me one minute.
[00:17:10]  Yes.
[00:18:01]  Yeah, sorry.
[00:18:02]  Yes.
[00:18:03]  So yeah, you graduated in 25.
[00:18:05]  Then so let's do the standard way.
[00:18:09]  Let me first give you a little bit of idea about me, myself, what we are doing, and then would love to know more about you.
[00:18:16]  And mostly today we'll be discussing more on the problem solving part.
[00:18:24]  You have done some good work on the assignment also.
[00:18:26]  So we'll take it from there.
[00:18:28]  If you have any questions about that, I'll be happy to take it and then we'll get deep dive into that.
[00:18:34]  So, myself Rahul, I have around 8 years of full-time experience.
[00:18:41]  I graduated in 2018 and then I have spent a couple of like out of these 8 years, 2 years I have spent in Bangalore working for a tech NGO.
[00:18:54]  We used to create platform and products for urban governments.
[00:18:58]  So, for example, your Hyderabad Municipal Corporation, right?
[00:19:02]  So they also use some sort of platform products to provide citizen services, tax collections.
[00:19:08]  So we created those for two years.
[00:19:12]  Then you've heard about Aruvge Setu in times of COVID.
[00:19:16]  Yes, sir.
[00:19:17]  I heard it.
[00:19:18]  Yeah.
[00:19:18]  So that was the project that our company was working on.
[00:19:22]  After two years, I joined a fintech organization in Mumbai.
[00:19:27]  Its name is Karza Technologies.
[00:19:30]  Karza Technologies was basically a company which had lots of API products for the banking and financial services industry.
[00:19:39]  For example, today, there are so many apps, right?
[00:19:44]  I'll just for an example, you say any there's a personal loan app where you just enter your PAN card, then it fetches all of your data and says whether you are eligible or not, right?
[00:19:55]  If you apply for any credit card.
[00:19:57]  So the company only takes your PAN card or maybe
[00:20:00]  mobile number and that's it.
[00:20:02]  It then gets pulls every data from GST portal from PAN card portal and then SS is the customer.
[00:20:10]  So they had lots of API products which were connected towards the official sources of country.
[00:20:17]  And then these API products were basically provided to companies like us, companies like, let's say, Tata Capital or ICICI Bank, right?
[00:20:28]  So that they can also assess the customer.
[00:20:31]  So I worked there for a year, then I had one startup of my own for a year in edtech industry.
[00:20:37]  Then I joined a B2B financing company in Delhi, like it was named as ProcCap, it was in supply chain financing.
[00:20:45]  So you have any idea, you have heard any idea about supply chain financing, B2B financing?
[00:20:52]  We will get into that, but do you have any idea about it?
[00:20:55]  B2B supply chain financing?
[00:20:58]  Not really?
[00:20:59]  Not that much.
[00:21:01]  Okay.
[00:21:02]  So I worked there for four and a half years and then I with a few of my ex-colleagues and a few more people, we started this company today in which we are working.
[00:21:14]  The company's name is...
[00:21:16]  NEEV Financial Technologies like NEEV like Hindi में जैसे NEEV लिखते ना?
[00:21:20]  Yes.
[00:21:21]  NEEV the foundation right that.
[00:21:23]  NEEV Financial Technologies is the name of the company.
[00:21:26]  Here what do we do?
[00:21:27]  So I take care of the complete technology part here.
[00:21:30]  I am the CTO here.
[00:21:31]  And what are we doing?
[00:21:33]  So, for example, we are partnering with lots of Indian brands who are distributing their products across the nation.
[00:21:42]  So, for example, let's say we are partnering with Samsung.
[00:21:46]  Okay.
[00:21:46]  Samsung has lots of distributors.
[00:21:50]  Pan India, right?
[00:21:51]  And then those distributors basically sell to small retailers.
[00:21:56]  For example, there are retail shops, right?
[00:21:58]  Who are selling mobile phones, mobile cover, etc.
[00:22:02]  So we call it a distribution channel.
[00:22:05]  So for every brand, one channel is...
[00:22:09]  The distribution side and one channel is the online side where let's say they are selling it on amazon their own website on social media etc
[00:22:19]  right so this is so we work on the distribution channel like where every brand has distributors on ground retailers on ground so that's our ecosystem in which we work
[00:22:30]  Now, what do we work?
[00:22:31]  So let's say Samsung has a distributor.
[00:22:34]  Now, Samsung gives them goods, right?
[00:22:38]  And Samsung wants money for its goods.
[00:22:40]  That's the basic structure, right?
[00:22:43]  Now, what happens in general trade?
[00:22:46]  So we call it general trade also general trade distribution trade.
[00:22:49]  What happens like the distribution distributor does not have the money on same day.
[00:22:54]  How will he have?
[00:22:55]  So if he is buying goods, he will also sell it to other retailers.
[00:23:00]  the money from them and then we he will have money to pay back to brand right in a very simple example
[00:23:07]  today every business works that way like you have let's say 50 lakhs of goods 50 lakh rupees to start a business as a distributor
[00:23:16]  you bought 50 lakhs of goods from samsung if you have the money for the first thing that let's say you will pay upfront when you buy you pay
[00:23:24]  Okay, but now if you need more, so then you don't have money.
[00:23:29]  So what you will do, you will ask Samsung that you give me the goods, I will pay you back in let's say 30 days and I will sell it in 30 days and then earn it and then give it back to you except my profit.
[00:23:42]  So that's how generally things work.
[00:23:44]  Now, but Samsung will say, I need money, right?
[00:23:49]  I need money because otherwise if everyone does not pay me for 30 days, how will I pay my factory owners?
[00:23:55]  How will I pay my labors?
[00:23:57]  Right.
[00:23:57]  How will I pay for raw materials?
[00:23:59]  Right.
[00:24:01]  Are you following?
[00:24:02]  Yes, I am.
[00:24:03]  Okay.
[00:24:05]  So there comes like company channel financiers.
[00:24:10]  What do we do?
[00:24:11]  So we say to Samsung that you give goods to this distributor and instead of distributor paying, we will pay you.
[00:24:18]  We will pay you on first day.
[00:24:20]  Okay.
[00:24:21]  So Samsung says that's okay.
[00:24:23]  My problem solved.
[00:24:24]  I sold the goods and I got my money.
[00:24:28]  And then we collect the money.
[00:24:30]  money from the distributor distributor after 30 days distributor will say okay i was giving money to you after 30 days now instead of giving money to samsung i'll give it to you no problem
[00:24:42]  yes okay so that's a very basic construct now why do we give money like basically we charge interest for 30 days the amount that we have paid we charge interest right basically if we have paid 100 rupees we'll take let's say 102 rupees
[00:24:56]  That two rupees is our revenue.
[00:24:59]  That's our.
[00:25:00]  that will help us in growing our business.
[00:25:03]  So that's the basic construct.
[00:25:06]  So we are based out of Mumbai and we are an early stage startup right now.
[00:25:10]  So more or less we are not a very big team.
[00:25:12]  It's a startup setup only.
[00:25:15]  We are a very small team of about 10 odd people.
[00:25:19]  And our tech team is also very lean.
[00:25:21]  Right now, there are like two, three people, four people actually.
[00:25:26]  So now we are looking forward for smart people who can come, who can build a product, utilize AI, how we can create different sort of products so that we can manage all these things.
[00:25:38]  Yes.
[00:25:44]  Yeah.
[00:25:45]  Now, mostly this is it.
[00:25:47]  And that's what we want to talk to you about now.
[00:25:49]  We'd love to hear from you, your journey, how it has been, what do you want to do?
[00:25:57]  Do you?
[00:25:57]  Yeah.
[00:26:03]  Yeah, sir.
[00:26:03]  The voice is breaking.
[00:26:08]  Is it better?
[00:26:09]  Yes, sir.
[00:26:11]  Yeah, so over there, we would love to know more about you, your journey, what have you done, where are you from?
[00:26:17]  Yeah, sir.
[00:26:18]  I am Tribhuvan.
[00:26:19]  I completed my B.Tech in Artificial Intelligence and Machine Learning from VVIT Nambusa.
[00:26:25]  Over the last year, I have been focused on software development, primarily working on Python, FastAPI, SQL, REST APIs, and full-stack application development.
[00:26:42]  During my training at Innomatics Research Lab, I worked on API development, database design, authentication and backend features.
[00:26:54]  I also built the projects like
[00:26:59]  fast uh sorry sir fast uh dentist appointment system booking using the mern stack uh and the self-driving loan onboarding agent which was the new assignment
[00:27:12]  uh and the job portal um what excites me um
[00:27:19]  me the most is building products that solve real business problems.
[00:27:24]  I enjoy understanding the business requirements first and then design a technical solution around it.
[00:27:32]  Yeah, that's all about me.
[00:27:37]  Okay.
[00:27:38]  Okay.
[00:27:41]  Let's just get into the assignment that you've done.
[00:27:44]  Like, what do you want to explain?
[00:27:46]  Like, what did you understand about the assignment?
[00:27:48]  How did you approach it?
[00:27:50]  What research did you do?
[00:27:52]  What are things that you liked about it?
[00:27:56]  First, let's talk about that.
[00:27:58]  Then we go deep.
[00:28:02]  Yes, sir.
[00:28:04]  The assignment is, when I first read the assignment,
[00:28:11]  I felt it was less about building a chatbot, sir, more about building and onboarding.
[00:28:22]  Your voice is a little breaking.
[00:28:25]  It will be okay.
[00:28:26]  I think if we turn off the camera, then you will see it better.
[00:28:37]  Yeah, sir.
[00:28:38]  Yeah, is it better?
[00:28:39]  Like, you are...
[00:28:41]  You can speak.
[00:28:44]  One second.
[00:28:47]  My charging is slow.
[00:28:48]  I'm sure.
[00:28:52]  Yes.
[00:28:52]  Yes, sir.
[00:28:53]  When I first read the assignment, I felt it was not building a chatbot, sir.
[00:29:01]  It was more designing an onboarding workflow for a financial product, sir.
[00:29:08]  The key requirement was that agent should not mainly hallucinate, sir.
[00:29:19]  or make assumptions about the customer data.
[00:29:23]  In a loan onboarding process, accuracy is very important.
[00:29:29]  Then generating the responses.
[00:29:31]  So before
[00:29:35]  I started the development.
[00:29:37]  I spent some time understanding how loan onboarding will typically work.
[00:29:43]  I looked at the KYC verification, eligibility checks and document verification and how financial institutions like
[00:29:56]  Credit B and other how it will work.
[00:30:00]  Based on that understanding, I decided to approach the problem as a workflow automation, sir, rather than a problem.
[00:30:10]  That's why I used the state machine architecture, sir.
[00:30:16]  So that agent guides the user, updates the forms automatically or dynamically, validates the information.
[00:30:24]  And if the data is not uploaded from the user side, it marks it as unavailable rather than showing, hallucinating some other numbers.
[00:30:39]  Okay, so if we have to create such a system like in the production,
[00:30:47]  then what are some of the things like suggest me three things which will break when we just take a workflow automation system in production where
[00:30:58]  State 1 comes before state 2, then state 3.
[00:31:03]  So if we take such a system in production, what are things that can break functionally?
[00:31:11]  Ya se va.
[00:31:16]  If we suppose if we take directly a linear state machine workflow, sir, as you said into the production, I think three major things can break, sir.
[00:31:32]  First, users don't understand the.
[00:31:37]  Sorry, users don't follow a consistent path, sir.
[00:31:41]  I mean the linear path.
[00:31:43]  In the real world, users can go back of the application and come back again after a few days, change
[00:31:53]  information they want or edit the previous steps like that, sir.
[00:31:58]  A straight state one, then state two, then straight three flow can be very difficult, sir, to manage unless we support backward transitions and like straight recovery, sir.
[00:32:13]  Second would be the...
[00:32:19]  external dependencies, I think, sir.
[00:32:22]  For example, KYC, PAN verification and bank APIs systems may be temporarily unavailable, sir.
[00:32:32]  In that simple workflow, if one step fails, the entire onboarding process can get halted at middle, sir.
[00:32:41]  In production, we need fallbacks.
[00:32:44]  That's the second one, sir.
[00:32:45]  And the third one is...
[00:32:54]  I think the predictability, sir.
[00:32:59]  The third one is.
[00:33:08]  Yeah, sir.
[00:33:10]  One reason I can choose the state machine or state machine.
[00:33:15]  No, no, it's mostly that's the idea.
[00:33:20]  No worries, no worries.
[00:33:21]  Let's move.
[00:33:21]  So the idea is same like state level machines are good when the workflows are simple and straight forward, right?
[00:33:31]  One after the another.
[00:33:32]  The only thing which breaks is the user might not have one thing after the other.
[00:33:39]  And if you want to switch any state, then how does it come back?
[00:33:46]  Then that has to be the intelligence layer.
[00:33:49]  It cannot be the state functionality.
[00:33:52]  But yeah, but great.
[00:33:53]  I like your project.
[00:33:56]  And for that only, why don't you tell me a little bit about any of the projects that you have done, like what you have mentioned in your resume?
[00:34:07]  Are there any AI related projects that you have done yet?
[00:34:14]  Yes, sir.
[00:34:15]  I work, I collaborate.
[00:34:17]  Yes, I collaborate.
[00:34:19]  worked on the code boss project sir it is an ai assisted review system sir
[00:34:27]  Which like at Enometrics Research Labs?
[00:34:30]  No, sir.
[00:34:31]  No, sir.
[00:34:31]  It was a collaborative project with my friends, sir.
[00:34:35]  Okay, what was the project?
[00:34:38]  It is an AI assisted code review system, sir.
[00:34:44]  Suppose.
[00:34:49]  The idea was to help the developers to get the quick feedback on their code before the formal review process.
[00:35:02]  A developer could submit the code, sir, and the system would analyze it and provide suggestions related to the code review, performance, and
[00:35:13]  like security, vulnerabilities, and all that, sir.
[00:35:17]  In that, we have used multiple agents, sir.
[00:35:20]  One agent for the...
[00:35:22]  How can you explain?
[00:35:23]  How have you used agents?
[00:35:26]  Yes, sir.
[00:35:28]  Yeah, I have used three agents, sir, in this project.
[00:35:32]  We have treated that agent as a system that can take and goal, gathers the required context and make decisions, sir.
[00:35:43]  For example, when a developer submitted the code, the agent first analyzed the code, sir.
[00:35:50]  It will identify the programming language.
[00:35:54]  Then it will evaluate the code quality, bugs and the optimization opportunities, sir.
[00:36:04]  How will it identify like is it a standalone project or you're talking about whenever someone is submitting a code in a let's say existing legacy project
[00:36:17]  there this agent will work.
[00:36:20]  How did so you created a POC or you created project like on what project was this agent used?
[00:36:30]  It was used in Codebox project, sir.
[00:36:33]  What is the Codebox project?
[00:36:37]  It is an AI-assisted concept.
[00:36:43]  It was a collaborative project, sir.
[00:36:47]  We worked on to help developers improve code quality during development, sir.
[00:36:55]  No, no, I didn't understand.
[00:36:57]  One second.
[00:37:00]  So, for how long this project lasted?
[00:37:02]  Like, how long did you guys work on this project?
[00:37:05]  Almost four, I think it was a long term project, sir.
[00:37:11]  I think it was, we have worked like six to five to six months, typically.
[00:37:16]  Okay, five to six months, good enough size.
[00:37:19]  Okay, so let's go bit by bit.
[00:37:21]  First is like,
[00:37:23]  So you created an agent and then you tested these agents on a project.
[00:37:30]  Is that understanding correct?
[00:37:33]  Yeah, sir.
[00:37:33]  Yeah.
[00:37:35]  Okay.
[00:37:35]  Now, what was the project on which you guys tested this agent that this agent works well?
[00:37:45]  Yes, sir.
[00:37:48]  The project itself was a code boss.
[00:37:51]  The objective of the project was to provide the AI assisted code review, sir.
[00:37:58]  Suppose where did you test it?
[00:38:01]  We have tested like our own GitHub link, sir.
[00:38:05]  Suppose we have cloned the GitHub repository and then we tested, sir.
[00:38:11]  Then you tested, okay.
[00:38:12]  Let's say you cloned any software and then you tested that you added a few more.
[00:38:21]  Changes and modifications to the code.
[00:38:24]  Now once you submit the code, then your agent kind of checks the quality of code and the impact that this new code will have on the existing code.
[00:38:35]  Right?
[00:38:36]  Is that understanding correct?
[00:38:40]  Yes, I am.
[00:38:43]  Okay, and how does it check the impact whether the new lines of air built impact?
[00:38:52]  How does it check?
[00:38:53]  Like what process does it use?
[00:38:57]  How is it able to or is it just that you have given the access to the whole code base and let's say you're utilizing a model?
[00:39:05]  Is it more, is it like a wrapper or is it like an agent?
[00:39:11]  Yes, sir.
[00:39:13]  It does not truly understand with the impact in the same way, sir, as a senior engineer would understand.
[00:39:25]  To be honest, it was closer to an LLM powered workflow, sir, than an autonomous agent.
[00:39:33]  Like the model was given the code and the relevant context, it will generate the review feedback.
[00:39:45]  Okay.
[00:39:46]  Okay.
[00:39:47]  Now, cool.
[00:39:49]  So, you mentioned like you were working, moving on, like you mentioned like InnoMatics Research Labs, you were on the training program for October.
[00:40:00]  25 to 26.
[00:40:02]  What was it?
[00:40:03]  Was it only what five months program or like it was it a full time job?
[00:40:10]  No, sir.
[00:40:11]  It was just a full full stack coaching, sir.
[00:40:15]  I mentioned it was a part of.
[00:40:17]  Was it a job or was it?
[00:40:19]  No, sir.
[00:40:21]  It was a training.
[00:40:22]  It was like a learning program, sir.
[00:40:24]  Okay, it was more like a learning program.
[00:40:27]  Okay, a course.
[00:40:27]  Yeah.
[00:40:28]  Course of the thing.
[00:40:30]  Yeah, sir.
[00:40:30]  Yeah, yeah, absolutely.
[00:40:32]  Okay, okay.
[00:40:33]  Got it.
[00:40:33]  Okay.
[00:40:34]  So, till now, have you worked with any company like any internship or that sort of thing?
[00:40:42]  Yeah, sir.
[00:40:43]  At InnoMatics Research Lab, I worked with a team of five or six.
[00:40:49]  It was overall like an internship, but it comes under the training program, sir.
[00:40:57]  A training program is okay.
[00:40:59]  There must be projects in which we were working on.
[00:41:03]  But have you had any internships before?
[00:41:05]  That's what I'm asking.
[00:41:07]  No, sir.
[00:41:08]  In your college, you did not try to work for a company as an internship.
[00:41:13]  Any specific reason why?
[00:41:17]  Yes, sir.
[00:41:23]  I did not do a formal internship during my college.
[00:41:27]  Most of my focus was doing AI-assisted code projects mainly and gain hands-on experience with different types of technologies.
[00:41:40]  I should have explored more internship opportunities during college, sir.
[00:41:45]  However, I spent that time building projects and strengthening myself, my skills, my foundations.
[00:41:52]  Okay.
[00:41:52]  So, Tribhun, like, what do you want to do?
[00:41:57]  Like, what is your...
[00:41:59]  Thought like currently uh so we are in mumbai and we operate from office right so what are exactly your priorities like currently
[00:42:12]  you're looking for a full-time job right now correct yes sir absolutely and you are open to relocate also
[00:42:20]  Yes, sir.
[00:42:21]  Okay, cool.
[00:42:24]  Take care.
[00:42:24]  Take care.
[00:42:26]  Any questions do you have for me about me or anything?
[00:42:35]  Yes, sir.
[00:42:36]  On which tech stack we would work more in a team, sir?
[00:42:41]  So there are majorly two tracks on which we are developing.
[00:42:45]  So one track is an existing platform that
[00:42:49]  we have that is built on PHP Laravel as the core stack, both front end and back end.
[00:42:58]  That is one.
[00:42:59]  And second is a new track, which is mostly about the AI related developments.
[00:43:08]  Creating the agents with that is on next.js and python so those are the stacks on which we are working
[00:43:18]  Okay.
[00:43:19]  And anything else you want to know about?
[00:43:26]  Are there any other rounds?
[00:43:28]  Are there any?
[00:43:30]  Other technical rounds?
[00:43:33]  Yeah, there would be a couple of more rounds.
[00:43:39]  Not exactly single rounds, but most problems solved.
[00:43:45]  Okay.
[00:43:46]  Yeah.
[00:43:48]  Yes.
[00:43:48]  Cheap.
[00:43:49]  Okay.
[00:43:50]  Yes.
[00:43:50]  Okay.
[00:43:51]  One, it was lovely talking to you.
[00:43:55]  Muskan and Muskan will be here next week.
[00:44:02]  Yeah, likewise.
[00:44:03]  Thank you, sir.
[00:44:06]  Thanks for having me today.
[00:44:08]  Bye.
[00:44:11]  Thank you.