Subscribe
Share
Share
Embed
‘Innovators Playground’ is a podcast series where we explore the cutting-edge innovations and insights shaping the future of technology. It brings conversations with leading founders, industry experts, and thinkers to inspire and inform youth around the world.
Dilip:
So today we are going to talk to Harshil on this whole journey of RazorPay, the last 10 years, how tech has evolved, how he has delivered almost uninterrupted services to the payment system. He's a large contributor to UPI and operates at a great scale in the market.
We are going to talk to him on AI. Some of the future trends, what he's tracking, how he himself is looking at being an AI enabled founder and a lot of many things. So, stay tuned.
Harshil, RazorPay is now 10 years plus and I could say that the UPI and RazorPay were born almost at a similar time. I remember the contribution you and Shashank made on the cool conceptualisation designing and the inputs on the UPI architecture. So, in that sense, I think it's great that two great things happened at the same time. If I have to just go through the 10 years, your tech stack, your technology advancements, would you like to throw some light on this?
Harshil:
Yeah, so first of all, on UPI, when we were starting RazorPay, at that time we were reading a lot about payments and all that. I remember reading the draft consultation paper on UPI and I looked at it and I was wowed. I was like, this is going to fundamentally change how payments happen. It was such a beautiful piece of paper because it borrowed ideas from everywhere else in the world and then said what is the best for India and it can be very technical. So I as a techie loved reading it because it went into protocols, it went into specific language to be used, specific consent mechanisms, authentication. Like it was really beautiful to read it. I remember reading it and I was like, if this really happens, it's going to just change. Of course, last 10 years has been a history on how that has happened. Going into our journey, right, 10 years back when we started building RazorPay, I think one of the earliest thoughts that we had is that we are building payments. And in payments, reliability is the most important thing. So what I can tell our techies is our business is like an electricity company. When it works fine, nobody cares, but if it goes down for five minutes, you would abuse it like anything. So nobody comes and says, hey, kya bijli aa rahi hai? (wow electricity is coming through?) Nobody cares about that. But 10 minutes if you're down, people are going to be really angry. So it's like a utility business and being up is the most important thing. It doesn't matter whether we ship the latest feature or if it's delayed by a week, if it's delayed by two weeks, but if you go down for a second in the hope that you' will be able to upgrade or do something better, people don't care what you are trying to do. So, one of the earliest focus for us was how do we make the system extremely reliable, and extremely resilient? We used the cloud infrastructure from day one. And while I don't spend a lot of time on tech side now, given the large team, but at that point of time, the first infrastructure of RazorPay, I set it up myself because I was on the DevOps side. So I spent a lot of time building the right kind of firewall architecture, building the right kind of database architecture to ensure that we can be extremely reliable. And over time we kept expanding on it, we kept building on top of it, but the core principle for us was always was reliability over innovation, even though innovation is extremely important, but reliability stays higher. So, for example, even today, we don't allow new features to be shipped on the core platform in the day.
Dilip:
Harshil, you know, I'm assuming and NPCI ourselves has been biggest beneficiaries of this whole open source softwares and the revolution happened since 2010 to 2020. And I'm sure RazorPay also would have been using that. How do you leverage the open source stack?
Harshil:
It's amazing in the tech world, the contribution of open source and the dependence on open source is so underappreciated at times, like so far companies like us, like the entire stack, the core stack is open source, the language that we use to the frameworks that we use to the most core infrastructure, something as simple as SSL that we use to drive securities running on an open source platform. And generally people realise it when it fails. And we realise that this is a small project by a small developer running, like sitting in Ukraine, who's been writing this code that your entire system is dependent on. And he's probably underfunded, is probably one developer writing all of it. And you don't know how much dependence you have on it. So, of course we have been big beneficiaries of it. And we have tried to give back a lot. We have a open source program in the org. We actually have open source Fridays every month with the engineering team where, we ask people to present, what did they contribute to open source and this team?
Dilip:
Wow, this is fantastic.
Harshil:
We also publicly posted out these are the contributions from our team in the open-source world and we encourage it. In fact, I'll give an early story. So, when we started using RazorPay, we were setting up for PCI certification.
Dilip:
Certification, PCI DSS certification.
Harshil:
One of the things needed for that is an Intrusion Detection Program.
Dilip:
IDS, yeah.
Harshil:
So we used OSSEC, which is a very widely used open source IDS. And I was the one deploying it. And I realised that there's some feature that is needed by most companies. There's a new SSL upgrade that had happened and it didn't support it. So actually we went and contributed back to that program and I'm still listed as one of the contributors to OSSEC in their GitHub repo. That's one thing I'm very proud of that I contributed back to open source in some ways in the early days. But that's the culture that we try to propagate in the team as well. That hey, can we spend some hours of our, every month of our Fridays, just contribute to open source and tell our teams. And we give people time that if you are contributing to open source on this Friday, you don't need to spend your time on work and we'll give you credit for that.
Dilip:
But when I look at the engineer's mindset, and that's very critical because our idea of this whole playground to how do we motivate the engineering grads to look at the startups and the FinTechs as a possible opportunity in their life after engineering. And when you encourage that from a company perspective, what's in for the engineer who's working for you? How does he look at it?
Harshil:
First of all, you have to understand the engineering's mindset. And I was an engineer myself. They relate a lot to it. Most engineers love to do stuff which have no really significant outcome. Engineers love to code. And that's one thing that, like it's hard for a lot of business leaders to grasp. Like business leaders always work on "hey, if I do X, I need this outcome. If I do this, I need that outcome." That's how business leaders are conditioned. The best business leaders in the world are very driven by what is outcome for me. Engineers are on the other side of the spectrum. So like I used to be used to code in college and I look back at the time and we used to code. We would like a lot of projects and spend our evenings, nights, 24 hours coding and writing projects. If you look back today, there was no outcome for those projects. There's nothing that I was getting out of it. I would sit in hackathons and yeah, there was generally a prize, but one person wins a prize, 100 people code. And today you have hackathons to do that. Some of the best engineers who like coding, like coding as a pass time. For them, coding is like playing cricket.
Dilip:
It's like a meditation.
Harshil:
Yeah, it's fun for them. So it's like, why do you play cricket? What's the outcome? Nothing, you're not going to play in a national sport, but you play cricket because you enjoy it. So for the best engineers and in fact, the best work, the most motivating work for engineers is when they are not chasing an outcome because then they are in control. So open source is one such area. That's one of the best ways to hone your skills. That's one of the best ways to pass time. And one of the best ways that you feel proud. So if you write a code, if you are able to push it to a public open source program and it goes in that, I know people in my college who used to spend their evenings coding on the main Linux repo and they would be really proud if they could get one commit of theirs or edit on the main repo because having your name seen in Linux contributors.
Dilip:
It's a pride.
Harshil:
It's a pride that you wear and you put it on your GitHub repo, you put it everywhere. So that's the engineering mindset. So the best engineers learn how to hone that skill and how to keep polishing that skill because if you start enjoying coding as a pass time, that's when you really spend your evenings, weekends doing and drinking and some of the best engineers in the world come out of that kind of behavior.
Dilip:
I think it's a great point. And today when we look and we are going to talk about the AI, I think this whole open source contribution community, right? And I think from an India perspective, while India has really mastered the art of services economy, I think this open source contribution can itself become a big kind of a job creator, and technically you don't need to be working somewhere to do that, right? You can start right from the engineering days or maybe even pre-engineering days to kind of do it. I think it's a great idea and I hope India...maybe there is some policy level incentivisation, education policy or something.
Harshil:
Awareness part. So, one of the things that I do whenever I visit my college and a lot of colleges is tell engineering graduates that hey, contribute to open source, because the recruiters today are looking at this.
Dilip: Wow.
Harshil:
So it's one of the best ways to make a profile stand out. Like there's thousands of resumes you get for any role. But if you can demonstrate to somebody that hey, I contribute to this, imagine if you can write in a resume that I'm listed as a contributor to the main Linux repo, your resume just stands out. So there is an outcome to that as well. Like there are recruiters today, there are tools that like from me like our CUs which try to look at people's GitHub profiles. And find out that these are the GitHub profiles that contribute a lot and use that to hire.
Dilip:
Fantastic. So I think the companies also can now look at how do they filter the talent based on the open source contribution.
Harshil:
Yeah, because that's a real world impact. Like if you're able to get your code committed into one of the largest global projects, then you know how to write good quality code. You know how to work with a large team. You know how to work with a distributed team. That's one of the best ways to demonstrate than any resume or any coding exam that you can give, that I can work with this in your team and contribute to it.
Dilip:
Fantastic. I think the really deep tech companies can actually adopt it like a first principle while filtering the talent. And actually then it can have a ripple effect on how the engineers are spending their engineering time. During the engineering, how are they spending time? What are they prioritising? I think it has a great impact on that.
Let's move to AI. And you know, NPCI also is going through a huge transition. We started AI fairly early. We built up the team under Vishal, who's our transformation head. But I think the one mistake I did is, we remained with AI, huge expertise on AI because the scale that we operate and the model. So how are you thinking about AI?
You know, I keep telling my HR that, please look at the way you spoke about the open source contribution example in the resume. Look at the AI tool example in the resume. Otherwise, you know, don't hire the person irrespective of whether he's working in product, HR, or anywhere, he has not worked with a AI tool. Then technically he's not a human plus AI. So how are you looking at AI?
Harshil:
I'll give an example. Like a lot of banks, when they looked at technology or when the internet or digitisation was happening, they created a separate team. That this is a digital team. They'll focus on digitising the bank. Everyone else continues to do the job. We now realise that’s foolish...
Dilip:
Yeah, I'm in front of you. I made that mistake.
Harshil:
So, most people realise because the true digital companies where every aspect is digital will always out-win somebody who is doing digital on the side. Same thing applies to AI. So if you put a team, "hey, this is the team that does AI, everyone else just does their job." So your business team does business, your marketing team does marketing, but there's one team that's coming and putting AI on top. You'll always have patchwork. You'll always have things on top. So you can do some fluffy stuff on AI, but true AI differentiation wouldn't happen. So AI is a fundamental change. Every single team in the org needs to adopt AI because if your org doesn't, somebody else will. And whichever org is using AI, it's such a fundamental shift. That outcome will be 10x of your org. You can't compete with that.
Dilip:
You can't compete on it.
Harshil:
Yeah, you can't say, "Hey, I have more resources "I can compete there." That outcome will be so significant. So first, every team has to adopt AI. And that's hard. It's like the engineering part is actually related.
Dilip:
So how are you fighting the mindset? Because I'm also having a similar trouble in NPCI. How are you hitting the mindset to, or incentivising them to move to AI?
Harshil:
It's multiple things. One is track every, first enablement. We have provided resources to everyone. So every engineer has access to Copilot today, which allows them to code, which allows them to use, because it makes (things) far more efficient. First you give that, then you do trainings. You train people, okay, these are the AI tools available today, this is how you can leverage them. And then you track usage. So we track copilot usage. We track, for example, we have reached about 30 to 40% adoption which I feel is very low. We need to reach like 100% adoption. So there are engineers who have access to Copilot, but you are using it to write only like 20, 30% of the projects. They're still doing this, because it's hard to change behaviors.
Dilip:
Hard to change the behaviors. Everybody's just typing the commands.
Harshil:
So you have to change that behavior. People are still searching on Stack Overflow and then copying code, versus like asking the AI to do that. So I think that's when you track usage and then you push people that, hey, like where do you feel the gap is? How can we re-train you? Can we do it? And I think that's when you drive behavior. The second thing that you spoke is on hiring. And it's not just engineers. For example, most of the business interviews that you do today, half of the questions that you ask in a business interview, people can answer with AI like this. You ask a chat GPT, it'll give you the best answer.
Dilip:
Perfect, yeah.
Harshil:
Don't look at it as cheating. You find candidates who do that well. Because on business side also, why do you want people who sit on the entire two days and create a PPT when AI can do it in five minutes? Five minutes job, yeah. So we actually have started adding that as a competency in the hiring process, where we look for people who are leveraging AI and using that to bypass the interview. Those are the best people. You want those kind of people. You want, yeah. Who will say, "hey, you need this PPT, I'll generate it in five minutes. I don't need to spend two days to write that PPT." So that's the second big aspect. But the third is again, AI will be something same as digital. It's a cultural phenomenon. You have to make people understand that it's a big change. And unless you are adopting it, you are getting obsolete. As a workforce, you're getting obsolete. As an individual, you're getting obsolete. As a company, you're getting obsolete. So one of the things that we do, for example, now is most teams host an AI learning session, where the teams ask people to come in and present what they are using AI in their job. And there's a lot of cross learning that happens in that because there are always people who are slightly ahead of the curve and who learned AI tools, they tell others, "hey, I'm doing this thing that we used to do, now I'm using AI to do that." One thing that I am doing myself is trying to use a lot of AI tools myself.
Dilip:
And the same journey I'm also--
Harshil:
Because unless you are-- same thing. You can't preach. If you are not at the top of your game, then you can't teach your team. You can't ask your team.
Dilip:
And technically, we are not the CEOs. I'm just saying we are one of them.
Harshil:
If you are not at the top of your game, you can't expect your team to be at the top of their game. For example, two weeks, three weeks back, I used this tool, Replit. I started using that. It's such amazing. I don't know if you've tried it. It's a programming tool, but it's not really like a programming tool. So it's an app. You download it. You tell it that, hey, I need an app on this. In five minutes, it generates the app. It generates the app end to end. It is so beautiful. I like to calculate the calories. And I asked it to create an app to track my calories. It created any other-- like my fitness pal. But then I said one step ahead. My biggest challenge is I don't want to enter each item. So at the end of the day, I want to send a voice note to you. Can you figure out from it and translate and note the things? And it integrated OpenAI. It asked me for the OpenAI key. It integrated OpenAI, generated the app. And then the app is something I'm using every day. And when you start doing that, then I was so amazed by it. I asked my entire leadership team that, hey, everyone needs to download this app today. Start using this today. Because then, for example, today, now when an engineer comes in and says, hey, this page will take two weeks today. But I can say, hey, I can generate this in five minutes myself. Why can't you try using it? So you need to try these tools yourself. Because otherwise, you don't know where the world is moving.
Dilip:
And I keep telling our people that, you know, or wherever I'm talking, that you are not competing with human now. You're competing with human plus AI. So if you don't have that AI into it, and whether it's a tool, it can be as simplistic as using the GPT, multimodal GPT, to do some of the rudimentary tasks you're doing. If you're not optimised on yourself, there are no takers for you. So I think it's fantastic. And you spoke about 30% contribution in coding. It's really good. How do you deal with the risk? Because today, there is a chance that the code might go outside, or the document might go. So how do you deal that inside the recipe?
Harshil:
Review process is still manual.
Dilip:
Review is manual.
Harshil:
Review is still manual. And we are not going to move. So, we can have one level of review as AI to find out if there's any, but the final review has to be manual. And so generally, in a coding environment, like large tech companies like ours, typically the SD1s, SD2s will write code. There's an SD3 of principal engineers who will review. And I'll treat AI as an SD1 or an intern level who is writing code. For my principal engineer or my senior engineer, it doesn't matter whether an SD1 wrote that code or an AI wrote that code. They still have to review. They still have to check quality. And so the review process is still fully manual. And for them, the filter is the same. So tomorrow, maybe half of the code that SD1s write is written by AI. My principal engineers and senior engineers will treat it like an SD1 and give feedback on it. And most AI tools are also evolving that they can take feedback and then rewrite code. So the review process continues to be manual. And I don't think we are going away from there any time soon.
Dilip:
Fantastic, fantastic. I think I'm going to shamelessly use some of the principles on AI, training AI, the appreciation and those kind of things for the people who are really doing good on that. As a founder, CEO, and the way you spoke about AI is going to make your game 10X, similarly, the talent is also the X factor for your success in that sense. So as a founder, CEO, how do you ensure that if there is a best talent and the RazorPay needs that best talent, how do you match it up? How do you attract that talent and how do you signal that you are the organisation where the talent should come and work for?
Harshil:
It's a hard problem. It's harder because the expectation of today's workforce is very different, especially on engineering side, than what it used to be. So it's not about who gives the most money.
Dilip:
Yeah, I guess so.
Harshil:
I think those are table sticks now. If you're a large tech company, you're hiring the best engineers, you have to give in the top 90 percentile. You have to be the top 90 percentile payer. You have to give the best ESOPs. All of that has become table sticks, honestly.
What matters today, for example, almost 60% to 70% of hiring today for us is still referral. And the reason for that is people like to find places where they like working and then bring people along with them. So the best engineers go to places where other best engineers tell them, hey, this is a good place to work. This is a good engineering culture. And that is becoming important because today, with the salary levels of engineering graduates and the top tier engineering folks, the delta between x salary level and 20% more is not significant. Their lifestyle is already the best it could be in a city like Bengaluru. So now they're looking for how do they work in a place where they like working. And that is a harder problem. So how do they find meaning in their job? And that comes a lot with empowerment and the ability to take decisions. So if you're treating them like a servicing company, that, "hey, this is what you need to build and just build it", then A, that engineer won't enjoy it. And secondly, like I just said, like tools like Replit is going to take that away. You're going to ask a tool and it will generate. You want engineers who are going to put their effort in and find out the small, small things that differentiate you. And that will come in when people enjoy when you're giving a problem statement to them versus just build this out. So we involve engineering in every step of the decision making process. Our engineering teams are very deeply involved. So in our business, for example, when we are discussing, hey, we are facing customer support issues and we need to build a better product around it. Our engineering teams are in the room. It's not like the business teams are taking all, hey, this is not what we need to do, solve customer support. We decided to create a spec, gave it to prod or prod created an engineering spec and gave it to engineers. Engineers are just coding. If you're using engineers like code monkeys, then you will hire code monkeys. So you have to use engineers as smart decision makers who you give a problem statement that, "hey, we need to find the best way to do this" and trust them to find the best way to do this. And then you will have the best engineers. It's a self-fulfilling outcome. A lot of companies mistake it because in most companies in India, I've seen a very business led where the business decides what needs to be done. Engineering is just an execution team. It's a servicing team inside a product company.
Dilip:
This is like, again, self-reflection because I keep talking about that, the engineers must be involved in every business decision because they are the one who are building it. And if they don't know the real user experience, they won't be able to build the craft. Rightly said that if the engineer is looking at an impact, one company may not be able to, or one work may not be able to give him that impact satisfaction and he's going to fragment his time. The future of workplace is going to change. So how are you thinking about the future of work or the future of the talent, which is going to come back to us?
The Gen Z as we call it, the Gen Z workforce is very different. And I think there are a few things that differentiate them. I think for the generation before me, work was like pious, responsible thing that you have to do and you have to go ahead at nine and you have to come back at five. I think in our time, our workforce, in our generation, it changed a bit where you work, but then work is work and then you have play and play and then work-life balance as a concept came in. Gen Z is very different. They're even more casual about work that for them, work is a way for them to enjoy their life. It's not that, hey, you work and then whatever time is left, you enjoy. You work so that you can enjoy your life. And I think that mindset first is very different. And I think it's a good thing. People are moving through that direction, right? Where you work, but then you travel. You work and then you spend a weekend. You work and then you go out. And I think that differentiation is going to become more and more important. And if you can't understand that, then you're going to lose the best workforce because the best folks will have that option to find the places that respect that. So we spend a lot of time on doing that. And I think that's why a lot of new age workforces have a lot of fun elements, have a lot of collaboration elements built in. So you say, okay, hey, you don't want to work like eight hours at a stretch. You are more productive when you work for two hours and then you go out and play TT. And then you come back and maybe work for two more hours. We'll provide all of that in the office. You have all of this available. You want to step out, come back. We don't really care. We don't care when you swipe in your card or swipe out your card as long as you're delivering the outcome. So it is going to become more outcome driven work. Hey, we want you to build this or we want you to solve this problem in this timeframe. And then you take a call when you want to do it. You show up one day a week and are still able to deliver it. You show up five days a week and unable to deliver it. Like you are rewarded that way. So it's going to become more and more outcome driven which is going to be harder for managers to imbibe. I think the workforce is an easier time. So because for managers, it's just harder to track that. It's easier for a manager to say, hey, you are in office seven days a week and five hours a day. It's easy for me to say, hey, I know you have done your job. Versus when you make it outcome driven then you do track the outcome. Is the outcome quality enough and all of that. And that requires the managers to be more informed, more aware of how much effort does it actually take to do that. And that's the reason most lazy managers go back to things like I want to track your working hours. I want to track when you come in. Because lazy managers don't want to put that effort to understand what it takes to produce this outcome and then push your teams towards that outcome. So the smart managers, so first is if you want that kind of the smart workforce, you need smart managers. And smart managers are ones that who can get on the ground, who can question our team and say, "hey, why would this take two weeks? That can be done in one week because I know how to do this in one week. So I can tell you." And then when you come to that kind of manager, then you have the best talent. So you need smart managers who are not lazy. And then you get the smart workforce who can work in that kind of environment.
Dilip:
And also change the KPIs. Look at the contribution on how much is the quantity of code written, quality of code written. Or the problem solved, something like that. I think the KPIs might have to change.
Harshil:
Yeah, so we call it touch points or the points of completion. So how many points of completion did you put out? How many outcomes did you generate? And as long as you're generating that, if you're generating more outcomes than the other guy, it doesn't matter whether you were in two days in office and this guy was five days in the office. The outcome has to be commensurate. So the workforce, I think the long-term view I have is that the best workforce will become very outcome focused, especially in the world of AI. There's an engineer who leverages AI and delivers it quickly. The engineer who doesn't leverage AI and sits five days and does it. Which kind of engineer do you want? You want the AI guy. So the outcomes are going to be more determinate of performance and I think we're going to move towards that. But at first it requires managers to understand this and change their mindset. It's easy for a lazy manager to say, hey, I'll just track what I know how to track.
Dilip:
Well, Harshil, while we spoke about the tech talent and the tools, the culture a lot, do you also work with some of the tech startups, and what's your engagement model with them? How do you create value for them or how do they create a value for you? What's your collaboration with the younger startups in that sense?
Harshil:
I think one of the best ways to learn what is happening in the ecosystem is to interact with startups. A lot of companies can do a lot of innovation, but what's happening in the world, where the world is moving? The best way to learn that is just to speak to early stage startups. Because they're the ones coming with the disruptive ideas. So that's the reason I started.
Dilip:
And they don't have resources. So they are fighting that battle, right?
Harshil:
Most of the newest startups that are coming up today are AI first, truly AI first.
Dilip:
Truly AI first.
Harshil:
Companies like us are adopting it and trying to become AI first, but they're AI first.
Dilip:
Likewise.
Harshil:
They're five people, they're doing 100 people outcome because they're doing AI on day one. So for me, I think that's why I started doing personal investment on the side. The purpose was not the monetary outcomes.
Dilip:
But that whole learning.
Harshil:
Yeah, startups are not the best way to make money. But the idea was that I get to interact with a lot of startups, I get to understand what they're doing, I get to question them, I get to mentor some of them and then see how they're doing things. And there's so many learnings I've got by talking to these startups. Because I talk to these early stage startups and look at how they're delivering such amazing outcomes today, early stage startups are doing so amazing outcomes with two people. With three people. Look at that and you feel hey, like What are we doing? Even with thousand people.
Dilip:
It's like sleepless nights.
Harshil:
Sometimes you look at that and say, he's putting as much outcome as my company is. And that makes you re-question your understandings and re-challenge the beliefs that you had. And I think that's one of the best ways that I do it. And I think every startup founder needs to do it. Like that's the best way that you can keep questioning your understandings.
Dilip:
So what's your future tech strategy? One is the AI, right? I'm just saying that every CEO is now thinking on whether the company gets AI enabled, the company gets AI-fied fully as you spoke about. But other than AI, what are you thinking, what are your tech plans going to be? Or, I'm just saying that while optimisation and efficiency is one of the objectives. But how do you see the tech trends playing around worldwide which are going to impact or enable RazorPay to do more?
Harshil:
Tech is a world that changes so rapidly. And especially in today's environment with AI, it's changing so rapidly, it's hard to keep up. And as I said, like one of the things that I try to keep doing is how do I use these tools myself to ensure that I'm aware of where the world is moving. It's just hard to keep up. So AI is of course the biggest trend and it's going to be the fundamental driving force.
Dilip:
We just completed the NVIDIA's four-day conference and every day morning something is coming to you.
Harshil:
Something new is coming, yeah, the Newton announcement.
Dilip:
The Newton announcement, yeah.
Harshil:
So first is like just, as a founder, as a CEO, you have to spend a lot of time in understanding where the world is moving. Because whatever I say today, like three months later is going to be a very different world. And like it's not moving at a pace of years, now it's moving at a pace of weeks. And like tomorrow, OpenAI will launch a new model which does 20 more things that you thought couldn't be done by AI for next two years. And that's the pace at which the world is moving. So AI is going to the fundamental driving force, but beyond AI, I think AI applied to specific solutions, specific problems, is still yet to come in. So we are seeing generic AI.
I think whenever a new technology comes, the first use case is generally very simple and lazy. So today people are just using chat GPT to ask basic questions. Hey, how do I do this? Hey, how do that? So it's a replacement for Google search, essentially. The deeper problem statements with AI are yet to come in. So you will see AI applied to, for example, insurance in a deeper way into underwriting, into fraud detection. We are seeing a lot of that. For example, we recently launched a buyer protection program which uses AI to identify fraud. Because when a fraudster buys something on a website versus when a real person buys, the way they interact is very different. And a human can't notice it, but there's micro differences. Like a real person spends a lot of time understanding the pricing and features, fraudsters don't care. These are things that AI can do much more smoothly. So how do you apply AI to solve this? You spoke about anomaly detection. How do you apply AI to solve and figure out common behaviors and predict that this is not a common behavior and it's failing. So that kind of like the specific use cases and applying AI to specific use cases and solving the problem in a way that a human couldn't is going to meaningfully change the world. So I'd put it this way that the cost of intelligence has gone down to zero. So anything that needed intelligence to put in, you'll say, who will sit and do this?
Dilip:
Yeah, just use it. Just pull in the expertise. Yeah, you pull in AI. AI is like a tutor, you're expert.
Harshil:
So for example, if I look at fraud management, if I could have a human watch every single transition happening, I would reduce fraud. But you can't do that today. You can't put a human today at multi-billion transactions that a UPI does. You can't have a human. But you can have an AI now look at every single transaction which is equivalent to a human and look at every single transaction and find out, hey, is this a real person? Is this not a real person? Is this a fraudster? And do that. So the cost of intelligence is going down. The way you have to think is that if I had infinite number of humans available and I could put them to everything, what else would I solve? Because that's the intelligence level that available to you now. And then you apply that framework. There's so many problem statements that you can solve if you could put a human to everything that you do. And that's what AI is giving you as an API now.
Dilip:
So Harshil, let's do some fun, rapid fire.
Your first programming language.
Harshil:
Oh, I don't know a lot of engineers won't like it, but I used PHP.
Dilip:
PHP, mine was Pascal. I'm more ashamed to speak about.
One startup that you really admire.
Harshil:
Globally, I'd say Amazon. Execution machinery that Amazon has. And like combining the business strength and tech strength both. Very few companies are able to do that. Like they're extremely good at tech, extremely good on business. And Bezos is something.
Dilip:
You know, it's a very inspiring choice because I've read one of the books on Amazon and there they do apparently million change managements every month. So even the change management works.
Harshil:
Every aspect like the shareholder letters that Jeff Bezos has written in the past, like those are all so much to learn from those. So it's amazing.
Dilip:
Very inspiring.
One book which you believe the aspiring entrepreneurs to read.
Harshil:
Yeah, very interesting one that I would say is Founders at Work by Jessica Livingston. It's a story of a lot of founders like Apple founder and so on Steve Wozniak and so on, their journey of starting up.
Dilip:
Journey of starting up, yeah. Fantastic, fantastic. You know, I have not read that.
A good habit that you have built last few months or maybe a year or so, one of the bad habits you're trying to break.
Harshil:
I'm trying to focus on my health a lot more. And I think when you're building a startup, you're in the start, especially in your young time, you just forget how important health is. And as you start to realise, it's not about focusing on health, like you don't want to build a gym body, but the focus on health comes because it impacts your productivity. And you start noticing the fall in your productivity because of the amount of effort you have not put into the health or the food that you're eating. So really trying to build that in the last few months, I'll say, but hopefully I can get it.
Dilip:
Yeah, one piece of advice you would give to your younger self.
Harshil:
You know, the one that I'll give is, especially in today's day and age is, be bolder. I think the world is changing faster and the things that you couldn't question, you can just do things. Like that's one of the common mottos, don't think, hey, how can this be done? You can just do things. Like there are people doing supersonic jets. There are people building anything today. Like you can just do things. And that's one thing I really love about Silicon Valley, people just take bold bets. If you look at those ideas and you're like, how could somebody even try this?
Dilip: Try this, yeah. So let's have a thinking. You don't need so much thinking.
Harshil:
And I think as you age, it becomes harder. Because then you have no so much, you say, hey, how can you do this? It's just so much effort. The best, that's why the best founders come from a very young age because you don't think how hard is it really? You say, how hard can this be? How are you going?
Dilip:
Fantastic.
Thank you so much, Harshil. I'm sure that many aspiring engineering folks will get inspired by this conversation. And also kind of look at your work or the work your team is doing. And also look at their career, pathing, charting along with the RazorPay or the new age fintech startups to make an impact, to create a really great value for themselves and the country in that sense. Thank you so much.
To the viewers, thank you so much for spending time with us. And I'm sure this has been valuable to you guys. And please do let us know your feedback in the comments section and stay tuned for many such exciting conversations. Thank you.