Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
4 Near-Death Moments to ₹1,000 Crore Valuation: IDfy Founder Ashok Hariharan on Survival & Scale
89 Plays3 days ago
In this episode, Ashok Hariharan, Founder and CEO of IDfy, shares the raw, unfiltered story of building India's largest background verification and digital identity platform from scratch. From nearly dying with just two months of runway in 2013 when 12 employees chose deferred salaries over leaving, to surviving a 90% revenue collapse during COVID by betting everything on Video KYC technology built three years early, Ashok reveals how patience and strategic readiness beat blitzscaling.
He discusses IDfy's evolution from a ₹3.5 lakh background verification startup to a comprehensive RegTech platform processing 65 million verifications monthly across onboarding, fraud detection, and DPDP Act compliance.
Ashok shares contrarian insights on incremental compounding over spike growth, building a 15% ESOP pool (largest in Indian tech), and why IDfy doesn't have "founders" but a leadership team designed for 40-year longevity. He unpacks India's hidden ₹10,000 crore fake employment industry, the technical architecture behind handling 100,000 requests per second, and why contributing to India's privacy law in 2018 positioned IDfy to dominate the DPDP compliance wave.
This candid conversation with host Akshay Dutt covers everything from rewiring the entire platform in Elixir during Diwali, to expanding internationally with 15% revenue now coming from Philippines and Indonesia, to the cultural philosophy of "Saraswati over Lakshmi" that shaped IDfy's approach to wealth distribution and organizational design. Whether you're a founder navigating the funding winter, building in RegTech or fintech, scaling background verification or KYC solutions, or simply fascinated by resilient startup journeys, this episode delivers actionable frameworks on manufacturing luck, surviving near-death moments, and building sustainable profitable growth in India's digital identity ecosystem.
#AshokHariharan #IDfy #IdentityVerificationIndia #BackgroundVerificationIndia #KYCSolutionsIndia #VideoKYCIndia #RegTechIndia #DigitalIdentityIndia #FraudDetectionIndia #DPDPActCompliance #IndianStartupJourney #StartupFundingIndia #FounderThesisPodcast #AkshayDutt #IncrementalCompounding #ManufacturingLuck #GigEconomyIndia #FintechIndia #AadhaarVerification #UPIFraudPrevention #IndiaStack #PrivacyComplianceIndia #CrimeCheckIndia #SyntheticIdentityFraud #StartupResilience #ProfitableStartupIndia #B2BSaaSIndia #EnterpriseTechIndia #SEAExpansion #PhilippinesStartup #IndonesiaFintech
Disclaimer: The views expressed are those of the speaker, not necessarily the channel
Recommended
Transcript
00:00:00
Speaker
He tells me, hey Ashok, what are you doing today? said, I'm running ID5. And he goes, Abibi, why are you running that stupid company? It's not going anywhere. Background verification is such a shitty business. What the fuck are you doing? There's a thin line between perseverance and stupidity, right? I was at the edge of that, at that point. It's fascinating how you have constantly been manufacturing luck.
iDeFi's Role in Identity Verification
00:00:20
Speaker
Ashok Hariharan is the founder of iDeFi. iDeFi has built India's trust infrastructure by verifying identities, detecting fraud and ensuring privacy compliance. Companies like HGFC Bank and Zomato use them for KYC and hiring decisions. We wanted to answer four questions, right? Does Ashokaririn exist? Is he the one doing the transaction? Has he committed fraud in the past? Is he likely to commit fraud in the future? The company is a fake company and there are 100,000 these in it. Your ID card, your PAN card, your driver's license, etc.
00:00:49
Speaker
There needs to be a way to... It's one of those traditional services businesses. None of your peers would be loss making. How did you burn through Entelax?
Ashok Hariharan's Entrepreneurial Journey
00:01:07
Speaker
Ashok, welcome to the Founder Thesis podcast. You're the founder of IDFI, which you've been building for a decade and a half, and you're also a serial entrepreneur. um I'd love to learn about your journey, ah both into entrepreneurship and of building up IDFI.
00:01:22
Speaker
How IDFI got started, a lot of it is, I think, some things are childhood related, some things are from high school, e etc. um So, yeah, I'm... ah i I was born in Delhi and ah very, like very early in my life, moved to KGF, Kolar Goldfields. If you remember that movie, that's where I grew up my first eight years. My father worked for the gold mines there and ah went to Andhra for a few years, like four years. Then actually did my high school in the Caribbean ah you know in a island called Curaçao. You've heard of this liquor called Blue Curaçao. That's where it's. Yeah, yeah, yeah. So then I went on to do my engineering from from Penn State in the US undergrad. My parents came back to India. They dropped me in the US and came back. um
00:02:13
Speaker
And a lot of what um the the kind of person I am and and and and ah ah And also, like my philosophies are kind of related to what happened in the US as well as what happened Warsaw largely because that's kind of where I think my personality got developed as what I am today. I'm kind of...
00:02:35
Speaker
the quintessential engineer, if you may. I worked in the semiconductor space host ah post my graduation, large largely building high-speed network processors in my previous life. I never thought an MBA was that important and never felt it it's ah it's a to be honest, a degree worth having. But but you know one like you know my parents one day called me and said, hey, um we we should get you married because you're earning quite well. And I said, theyll I'll go to my MBA and and last attempt at love marriage, as you may. And I didn't find my wife at B-school. So... Congratulations. So, um you know, ah I think achieved what I wanted to out of my MBA, but I'm still, I think, heart of an engineer, think like an engineer. Initially, when we started the first first business that we I started was something around EdTech and we wanted to do what Coursera was doing.
The Pivot to iDeFi
00:03:30
Speaker
And i'm I'm more of a tinkerer and a product guy. So like the product part of it was more exciting than than i would say the business part of it in those days. right and and And that's kind of how IDFI started as well, is is deep product thinking is what I would put it as. but two When did you ah feel like the edtech was not working? There is time to shut it down and not just have resilience. You know, resilience is like one of those things you hear as a... I didn't shut... I mean, ah me and my business... i mean, me and Vinny Chawab, we started both companies together. he he ran Gaboli, which is the what we call that company. In 2008, we started that.
00:04:10
Speaker
two thousand and eight we started that What happened is we see money was not available. So we kind of ah ended up um doing like, you know, alumni management software, and then we we started making money from it. but But eventually it became a services company because there was no way ah that that the idea of Coursera was maybe too early. And and when that happened, i that was not my cup of tea. I um um i kind of wanted to do products. So I ended up um And then we had this idea of IDFI and I said, hey, I'll run this and you run that.
00:04:42
Speaker
And so so that's kind of how it it all happened. But Gaboli, Vineet ran it till 2017, almost 2017 or 2018. As a services business. As a services business in the end, it shut down. but um Yeah, so ah like really honestly, it was ah it was about I'm a product guy and I wanted to do product. Like I said, it wasn't about the business as much as the idea of ah of solving a problem, which is big enough.
00:05:09
Speaker
In 2011, basically when Gapoli started and we posted a job on Naukri, this was in This was in 2009 or 10. And three people applied. And in those days, 1,600 people would apply to any job. Nowadays, it's almost very difficult to find people. But in those days, it wasn't the case. the word So 1,600 people applied. Three applied to the same resume. word for word, including grammatical mistakes. I will be working at Infosys for the last three years. And we called them, right? We called them over for an interview and we realized obviously they were lying. And that kind of triggered a thought that, hey, you know, ah maybe India, there's there's fraud and can we catch fraud? And that's kind of where it, like that conversation started between me and Vinit saying, hey, how do we catch a catch a fraud and things like that? And in fact, when we started, we wanted to do what um
00:06:02
Speaker
what DigiLocker is today, which is self-sovereign identity, um that and you wanted to to wanted it to be a consumer business where a person can, like, you know, let's say um let's say you're you're your, your Ashokariran, but you want to only like authenticate your age. Why should I give my whole ID card, right? The idea was, I just should be able to say, hey, my age is authenticated. That's it, right?
00:06:24
Speaker
um So, so, so that was the idea behind ID5, but obviously that didn't go the way it did, but we became more enterprise-y, but the original, original idea was a self-sovereign identity with full control on the consumer side um is how it all started.
iDeFi's Early Days in HR Verification
00:06:38
Speaker
and And that was exciting in those days, right? Like ah think think cryptography, you know, we were thinking, okay, how do I how do i make sure? How did you ah launch it? like Like, did you launch a consumer version or did you go to businesses or like, how did you discover that PMF?
00:06:54
Speaker
Yeah, I think, ah ah again, 2011, right? Not too much VC money available in those days. I think Flipkart had just raised about $10 million. dollars um So um we started with just enterprise. And in those days, only there was only one industry where verifications would even happen. That was the HR industry where background verification would happen, right? So so we started there. Pretty old school sort of business. um Almost all my competition, actually all my competition at that time were still doing it with paper, pencil, Excel sheets and things like that. And we just put a form together and and started it in that fashion, right? Like just just an online form literally. and And just then create a small workflow for for operations team to handle that. And that's sort of how we started. um
00:07:43
Speaker
And there was no other way to do it because um you know you we didn't have the kind of money that that floats around nowadays, right? ah So I raised about 80 lakhs in my first round from Bloom and a few angels. and Which year?
00:07:58
Speaker
2012. Okay. okay So we made... this ah after some After PMF or pre-PMF? like um Just pre-PMF. ah Because... Yeah, it was pre-PMF because Bloom was doing early seats sees ah seed stage checks in those days.
00:08:17
Speaker
And so, yeah, like, I mean, and then we we we went after the HR world, which, so we started generating revenue on almost day one of our business. So I think the first year we did about three and a half lakhs in revenue. It sounds very odd to even say it, but so that, that like you know, that's sort of how IDFI started.
00:08:39
Speaker
ah So this is like a pre-employment verification yes service. Yes. ah yes What all were you verifying? like So it was not not automated like we are today or anything like that. It was just um it was just a better way of of way of doing the verification, I felt.
00:08:57
Speaker
um but But on day one, we had we had already like beaten all all our competition because all of all the rest of them were still doing it with forms and, and you know, on handwritten physical physical copies and things like that. We said, hey, why don't you upload it here and and we'll do the ah like you know operations optimization where like you own the platform would automatically hive it off to specific agents rather than rather than a lead handing it to a specific person. So we were able to bring down the average, like, you know, that from like 27 days to like seven days. That was a big win.
00:09:34
Speaker
ah is right So that's sort of how how it all started. And it wasn't wasn't anything ah spectacular. It's just that my competition was just horrible.
00:09:44
Speaker
and the how like How do you get ah verification of a degree, for example? like because we We send a a letter to the, in those days, you would have to send a physical letter sometimes because emails they wouldn't respond. So sometimes they would respond with email. Sometimes you would send a physical letter and they would respond back. But ah ah nowadays, obviously, a lot of things are automated now. But in those days, it wasn't.
00:10:10
Speaker
And that is, hey, you know if i have if I have education tasks and employment tasks, I can hand it over to separate agents. I don't need the lead. So we just kind of optimized the process parts of it. and And we just killed some pieces of that process, which allowed us to speed up significantly. That's it.
00:10:25
Speaker
but but but yeah This sounds like a people heavy business. Like you would need a lot of feet on the street to do address verification and stuff like that. It was. In those days, like that's how we started. And over a period of time, we kind of optimized it. Like in the sense, I think um we were...
00:10:45
Speaker
We were probably, from a people in intensity perspective, we needed half the number of people that other BGB companies needed. needed So just operational efficiencies, ah because we put tech behind everything.
00:10:57
Speaker
um And we built our own workflows and things like that. But yeah, it was it was a very ah people heavy business ah still, even even though even though we probably needed less number of people.
00:11:10
Speaker
Okay. Okay. And today, what percentage of your revenue comes from this? The BGV? About 15%, 20%. Oh, wow. Okay. So 80% of the business was built. Yeah. And honestly, like, you know, um the BGV business is very close to my heart because of the fact that that's how we started. um but But today, IDFI is a lot bigger, a lot different than what we started um out as. and and we evolve and And I keep talking about this quite a bit is that um There's a lot of value for ah incremental compounding.
00:11:44
Speaker
ah People think that you know you there are these moments where you'll have this spike and stuff like that. But if you can just keep keep improving and keep incrementally adding value, um over a period of time that compounds. And and ah and I don't think ah people give enough value to incremental compounding in in my opinion. And IDFI is a perfect example of incremental compounding. right Start with BGV, then move towards what we are today, which is a three platform story around onboarding, risk and privacy.
00:12:13
Speaker
And that's not where it started from. But if I were to take any moment, ah I can't say how we evolved. like you know I can't take that one moment and say we evolved to this position in 2017. That's not how it works. All these little bits and pieces of decisions over a period of time, ah which gets you here.
00:12:32
Speaker
And gets you to that scale. And and with that, you get a lot of organizational memory, a lot of culture around being able to build incrementally. Right. now I think the spikes are actually, ah i mean, well, they're good, but you don't build ah organizational memory with spikes. You build organizational memory with with compounding effects. Right. yeah and And that's not given enough ah enough respect in India, I feel. While I think that's sort of how... how how you can build long-term enterprises in my opinion.
00:13:06
Speaker
ah So take me through the journey. So 2012, you raised round. You had money in the bank. You had some revenue. yeah Then what? So 2012, we had money in the bank. 2013, we kind of, end of 2013, we kind of ran out of capital.
00:13:23
Speaker
um And you know, when when when when you run out of capital, when things get tough, that's when founder issues come in play as well. So we we were we were three founders, my friend.
00:13:33
Speaker
ah current co-founder, which is Vinny Chava, me, and and another person. um And ah you know we obviously had founder issues and around vision, or this is not the direction we need to take, et cetera.
00:13:49
Speaker
um And, you know, when when ah when a founder co-founder leaves, ah no VC wants to touch you. um Because they start, and and you're subscale, right? You're not you're not not yet hit. um I think second year, and my revenue was 30 lakhs.
00:14:03
Speaker
So we have not hit any any any major milestone. the The jump is impressive, though. 3 to 30. I mean, it's nothing. no Right.
00:14:12
Speaker
that's like zero to a 30 yeah okay true um so we had about two months of money left i think um and uh and we were obviously not not profitable ah we had 12 people um i remember going to ah i called a i called a team meeting and i told all of them that hey you know what we have two months of money left i can If you want if you guys want to leave, I completely understand. And they said, what are you going to do? I said, I'm going to try to run this, but um I'm okay if you guys want to want to leave and you probably should, ah right? ah in all In all honesty and... and and and
00:14:49
Speaker
and those 12 guys, actually, I have four of them still in in the company, got up and said, we have to do it together? You know, all upon of that. and so And that was like, that was a moment which which hit me hard.
00:15:02
Speaker
And they said, don't pay us for whatever number of months until you find, until you get your, get the next round of funding. And we'll see how long we can last, right?
00:15:13
Speaker
And ah that hit home, um ah you know, TRE moments and all that all of that. But, yeah, I think that's that's one sort of really a scar which hit deep and and I think I still hold very dear to me, saying, you know, this is the kind of ah people that we have hired.
00:15:31
Speaker
and um And obviously, we we were able to raise a bridge around with Bloom about three months later. But I did ask my team afterwards, like, why? Why did you guys do this?
00:15:41
Speaker
And they said, you know, this is the best company we've ever worked for. We didn't want it to die. So, like, you know, that that kind of told me that, you know, okay, we are we are fighting a battle together. we're not fight I'm not fighting this alone.
00:15:53
Speaker
And ah a lot of them still are here. and And even we have our annual parties, they still show up, you know, those who are, and we call them the apostles of IDFY because 12 of them, right? so So they still show up there, you know, um and and very close to close to the company even today. um But yeah, that's i think I think part of the culture of ID5 and what we try to build is is is is deep respect for people. And I think that's what that's why that's why they stuck around. But that also allowed me to fight the battle many times over. We almost died like four times. So this was the first first near death moment, right? Tell me something.
00:16:35
Speaker
See, the background verification business is like, it's one of
Financial Challenges and Tech Investment
00:16:40
Speaker
those traditional services businesses. None of your peers would be loss making, right? These are typically businesses, lifestyle businesses, i could say there would be lot of, like say recruitment agents. I used to run a recruitment agency for a decade. So i like like these are typically not loss making businesses. How did you burn through Etilax?
00:16:59
Speaker
Yeah, I mean, ah so we wanted to invest in tech, right? So we had it had a largest tech team. ah my my None of my compet competitors had any tech team, right? Like we were the first ones even put tech behind us. Tech people cost money, right? That's one. um Second, I think, ah um um I mean, we it's 12 people. So we we kind of...
00:17:22
Speaker
ah 80 lakhs, think about it. It's not that much. It's like three tech guys, four tech guys and you're done. Yeah. Yeah. yeah but But do you regret that approach or do you think it made you who you are? Which approach?
00:17:40
Speaker
ah You know, like ah that overspending on tech. No, I think it it was a right decision. I still do it. I think um amongst my competition, even in the current business that we run now, which is a far more evolved as a company,
00:17:58
Speaker
I have a larger tech team than any of my competition does. I have ah i have a deeper ai team than anybody else does. ah um So I think we were very clear upfront, right? That the end all for us was not background verification. It was it was the fact that you know there's fraud in this country and can I solve for fraud?
00:18:19
Speaker
um That vision stuck around for a till even now, right? Like we we are fraud busters or we call it call it the ah trust stack of India.
00:18:29
Speaker
And if you're building a trust stack for India, then then I'm not, um and then I have to make that investment. And plus I'm a techie, like, I mean, right yeah Danda is not the reason i started at some point. that okay so mean Highly profitable, but but now, but it's the, I think it also, you know, there's a product market fit and a founder product fit ah as well. and And you can't take ah a product guy, engineering guy and tell him to do something else or or you you won't you won't fit in any other.
00:19:06
Speaker
Yeah, yeah, yeah. So, ah okay, Bloom gave you that bridge round then? Yeah. So ah Bloom gave me the bridge round 2013. Then again, we kind of... ah like got to break even by about just about. And 2015, how much revenue did you do? 2015, think 2015 would have been about crore, crore, crore and a half maybe.
00:19:33
Speaker
three sixty i think two thousand and fifteen would have been about a crow okay grow growing of maybe So we ended up raising from NEA at that time.
00:19:47
Speaker
um And just maybe six months before that, again, we had another one of those moments where we almost almost died. we We were just about, we didn't have enough cash, but we were just about break-even.
00:19:57
Speaker
ah And and and i'll I'll tell you one very very funny story, right? um In 2015,
00:20:04
Speaker
um There was this, me like one of the leading VCs in the country, I won't name the person. um um he He called me ah for a meeting and i I was very excited thinking like, oh, you know, this this VC is calling me. So maybe maybe there's there's some interest in what we are doing.
00:20:21
Speaker
And then this is like inbound. You didn't reach out. And yeah. And I mean, one and a crores is not a very big amount in those days. And and even even in those days, it wasn't considered to be a big success or anything like that. so So, yeah, he called me. So i was very excited. And I remember going to him and and I told the team that, hey, we got call we got a call and we go to meet this person.
00:20:46
Speaker
And he tells me, hey, Ashok, what are you doing today? I said, I'm running ID5. And he goes, Abibi. I'm like, why did you call me? Why are you running that stupid company? Like, i mean, it's it's not going anywhere. Background verification, you know, it's ah such a shitty business.
00:21:05
Speaker
what What the fuck are you doing? you know Sorry, but pardon my French. That's fine. And I'm like, okay, why did you call me then? He said, there are some roles in some of my portfolio companies.
00:21:18
Speaker
So ah as CXO roles, you know, you're a good product guy. Why would you be doing this stuff? And and he was obviously a well-wisher. I'm not... I'm not for a second like you know um but positioning as if like he did something wrong there but but um yeah he he told me that and and ah it hit me like I sat there and going oh shit you know have I like succumbed myself to a position where this company is not going anywhere I'm just i'm just kind of ah you know sitting here doing things which which I could have been doing something better, you know, because it's been three years, right? Four years and four years, I'm still at a crore and a half.
00:21:58
Speaker
um Like I can barely pay ah pay the salaries. i was not taking any salary. ah My friends are telling me, ka than dakaraya this is like a this is like, just like what you mentioned, right? It's an operations heavy you business. what Why the hell would you do this? You're ah youre a techie. why why Why are you doing the this kind of business? and and ah And everybody like I knew kind of said that.
00:22:21
Speaker
So, yeah, it's, it's, it's, uh, so you start to question like, am I doing the right thing? Am I, am I, uh, am I, am I crazy? You know, you're, you know, there's a, there's a thin line between perseverance and stupidity, right? Yeah. true true truth and know i was at At the edge of that, at that time at at that point.
00:22:40
Speaker
oh and ah and And, yeah, yeah so so long story short, you would sit there and go, you question yourself saying, at that point, maybe we had about 25, 30 people in the company.
00:22:53
Speaker
And um you're questioning, hey, you know, this does this make sense, blah, blah, blah. But what kept me going was still that same team. That team was intact. The 12 people are still there. And I felt like, hey, this this this I can't let this team down. Like, we are building something. a great culture. like it's you know We felt we are unique in terms of culture and we said, hey, we we should figure out other things to do because ah only background verification existed in those days. ah If you remember, the KYC process was still physical in those days. RBI had not allowed for digital KYC.
00:23:28
Speaker
So there was no construct of digital KYC. There was barely, Aadhaar had just started penetrating. There was no UPI. ah right India's stack was just basically Aadhaar. So you're sitting there going, ah you know um youre and we wanted to be the India Stack, right? And then you're sitting there going, is it happening? Is it not happening? yeah Am I going to be stuck in this background verification world ah forever?
00:23:55
Speaker
um We wanted to move towards the financial services side. Nobody cared about risk that much because actually there was not that much fraud in India in those days. ah Fraud actually increased post-UPI, if you may, post-digitization, that move. So 2015, again, you're starting to question, but you we were able to raise from NEA because, again, we started to think about, at that point, we started to think about hey what what would the world 10 years from now look like?
00:24:23
Speaker
ah ah how how would virtual like you know virtual transactions would increase and how how do you how do you then build build an IDFI of the future, right? We also saw the likes of Uber, Ola picking up. like So we started saying, okay, gig economy is picking up. There's a lot of scale in gig in the gig kickw world. Here, tech would make sense in order of for authentication, but nobody wanted to authenticate anybody, ah right? like People largely felt, hey, you know we are growing. We don't need, there's no risk.
00:24:52
Speaker
And if you remember, 2015, December is when um when that Uber rape case had happened. which Which led to the death of taxi for sure. as that' high Death of taxi for sure. um So we were we were shouting at the top of lungs saying, verify the drivers, verify the drivers. Nobody did.
00:25:11
Speaker
But that moment then just got that... god got that ah that you know that inflection moment for us where gig gi economy opened up. um And honestly, it was just the taxi guys in those days. it was no swggy There was no was no Somato or barely. Swiggy might have just started, if you may. And and and that's when that's when this the real wave of scal ah you know scaling issues, et cetera, started because we were doing at that time maybe 150 verifications a month.
00:25:42
Speaker
um And when when this... this ride-sharing world picked up, we had to all of a sudden scale up, and we were the first ones to get it, we had to scale up to 4,000 verifications a day.
00:25:55
Speaker
So you can't do that without without implementing massive tech, right? Like we had to automate the heck out of it. So ID card, ah ah yeah we had to start OCRing, so we built OCR tech ah on our platform, which kind of allowed us to just take a photo and do the full verification, et cetera.
00:26:11
Speaker
And that was the origin of our ah our scaling up story as well as the origin of some of the tech that we then pushed into the web banking side. yeah you You would kind of think that banking would be the first one to start and then you would move to the other side. But it's exact opposite. It actually started a white collar, then blue collar where scale happened, then the massive scale width with ah the post-Aadhaar and digitization world post-UPI world that happened. right So 2015,
00:26:39
Speaker
We raised about three and a half million because on the back of this gig economy ah sort of pitch that I had made saying that gig economy, we need, this requires a complete rethink of background verification because the traditional physical verification, you know, the operations heavy business had to change.
00:26:56
Speaker
And we came in and said, this is this is the new way of doing it. um And we moved from about one CR in revenue to about three and a half CR in in a year. So that was the first big shift, right? um okay how So BGV is ops heavy.
00:27:14
Speaker
um You often need to maybe call companies to verify whether this guy was working with them previously, e etc. etc how How are you running that with just 12 people? um So we were automating everything, right? like So we don't have to call a lot of times you just, ah you just have to essentially send an email that responds. So if I can automate the email, um and 75 80% would respond anyway. So what traditional BGV companies were doing was actually ah emailing one by one, we just said, hey, we'll automate that that part of it, let the first email go, a second email go, and then we will do the follow ups, right? And we would get 80% of the response would just show.
00:27:49
Speaker
come with an email. Same thing with okay and same thing we did in the on the on the university side as well. We just automated
Scaling Verification for Blue-Collar Workers
00:27:58
Speaker
printing of like the en envelopes so that the then we would print like 50 of them it's completely automated and then and then bulk send out um send out those letters so that we don't have to have an individual doing that.
00:28:11
Speaker
um So we we kind of ah did what what you would call assembly assembly line automation, really, not really any FADU automation, but just saying, okay, we we can do this through tech and simple tech, nothing major. Okay. The BGV business did not require any, like say, a address verification by somebody actually going and visiting. Like, I mean, the white collar site in those days, they never did address verification.
00:28:37
Speaker
ah But on the blue collar side, we needed it. um And when the blue collar side, as ah in 2015, when we started that sort of when that when that idea started to pick up, we we had like just like ah what Swiggy or Zomato or Dunzo had in those days, we had a app where people would get tasks and we would pay on a per task basis.
00:28:57
Speaker
So there were three, one point, and even today for for that matter, there are about 4,000, 5,000 agents um who get who get paid on a task-by-task basis, but they run they run their tasks on our app. So we built an app which they use even today, by the way.
00:29:13
Speaker
You built your own gig platform in a way. Yeah. Okay. So by the way, in 2015, that was a pivot that we wanted to do as well. Because we realized BGB was not scaling. One of the things we brought up to the board was, hey, we would want to do delivery, something like what Dunzo was doing, right?
00:29:38
Speaker
um like a task rabbit type thing. who done so And at that time, the board said, hey, stick to your guns. At some point, this is going to this fraud thing is going to play out.
00:29:50
Speaker
And Karthik from Bloom had told me that. And i like I don't know why you believe in it so much because I have given up at this point. wasn't scaling, right? Like only white collar was happening till then. And then and then until... But there were, I think, scaled BGV companies. Yeah, love me all of them were subscale in those days, right? Like matt the biggest guy was 30 crores.
00:30:14
Speaker
Okay. Right. ah Like, so you're... Was it because it was fragmented or employers were just not looking at this as product? There were not that many employers, no not too many people were doing... The TAM was small. Like, I mean, BGV, I think penetration was about, if I'm not mistaken, 10%.
00:30:30
Speaker
of the companies did BGV. Today, that number is maybe 30-35%. And also number of people see with the BGV business, it's about it's about churn. right like you you As you hire employees, you do more verification. So if if you have a lot of churn, then you would get more and employees to verify.
00:30:49
Speaker
But imagine 2015, nobody was leaving their jobs. It's not like today. so You know, you would only do those incremental verification when the company is growing, not really when when when there's actually imp employee churn.
00:31:03
Speaker
Today, employee churn is huge. Wasn't the case in pre-2015. pre two thousand and fifteen And you were serving enterprises, startups, what kind of clients?
00:31:14
Speaker
ah Largely enterprises, ah ah startups didn't care about BGV. BGV, true. And who did the sales? Like, how did you crack those accounts? And these would be sticky accounts, I guess, like once you get in, you're in.
00:31:27
Speaker
Yeah, i mean I mean, yeah, it was like large largely sticky and we had one salesperson that's about it. i And yeah, so so I think it was largely ah just me and him doing the sales and and the white collar, like I said, white collar wasn't scaling ah because of the reasons I mentioned is this. Penetration was not there. And number two, there was no employee churn at all or very little employee churn in those days.
00:31:53
Speaker
ah Until Flipkart started hiring like crazy and all the VCs started funding startups, I don't think there was that many that many people leaving their jobs. Yeah, yeah, yeah, yeah. True. Right. That's probably more of a 2015 plus phenomenon rather than pre-2015.
00:32:10
Speaker
yeah Yeah, I mean, I used to run a recruitment agency till about two years back and 90% of the recruitments that we would facilitate would not have a BGV thing in it. Although we were mostly working with startups. Yeah, startups in those days. ah but You know, in that process, if you think about it, like you you start to question yourself a lot because ah the other other parts of the businesses, like, you know, for example, Geek Economy had not picked up. Nobody wanted to verify ah The financial services industry was still old school, like coming and picking up paper from your house and things like that. So you start to question, hey, you know and fraud was not very big either. So, you know, what you thought that Indians lie a lot and therefore fraud will happen. It was not like really showing up in the market. So you start to question saying, hey but ta yeah mean like you what what what truth ah all the industries that I thought will need it as as ah don't need it yet.
00:33:05
Speaker
um And that's what 2015 going in was. But when that gig economy wave happened, um um I think that's when we got our funding and we started to grow. right ah Were you able to ah create like that repository of verified people that could be reused? like the same yeah We tried it three times. One initially um didn't work for white collar.
00:33:31
Speaker
Then we tried in 2015 with blue collar. That was one of the ideas that I had pitched in the in the in my fundraise. is saying, hey, why blue collar, these people are going to get employed in the in the gig economy world. Can I pre-verify them and give them an ID, like you know almost like a verified ID with physical ID? The idea was ah they don't have a calling card. Can we give them a calling card which they can say, hey, I'm verified here as IDified person? right In fact, ah we we ran that for about about six months and spectacular sort of
00:34:02
Speaker
spectacular sort of ah response we got from the blue collar side. ah We added about within about two weeks. And this was free for them. This was free for them.
00:34:13
Speaker
ah We added about ah about three lakh people onto this platform just in Bombay alone in about two weeks. Our attack was one rupee per person.
00:34:25
Speaker
ah And ah it's unbelievable. One rupee per person. How did we do it? We put ads on local Marathi newspapers and on best buses saying, um and know could clariify a low you know, that kind of thing. And and everybody wanted it, right? Because, the but but the problem is nobody was willing to hire them.
00:34:46
Speaker
So one side of the market was ready for it. The other side was like, hey, I like, and maybe I all, maybe in in retrospect that now that I think about it, maybe be the worst of the worst ended up joining this ah platform. the People were desperate for jobs. Yeah. And then these guys on the other side didn't den ah then't find value in those people. Possibly.
00:35:09
Speaker
But the demand on on the ah on the individual side was high. ah the The supply was good. that The demand side was really bad. Right. So so we and actually then killed that business in about six months. And ah yeah, it was crazy. It was ah we were we we thought we had won ah because it was like one of the red herring moments. And teen lark in two weeks is unheard of. Right. Like no one done that.
00:35:35
Speaker
ah But yeah, we couldn't we couldn't actually get the demand side going.
Adapting to Market Demands with Digital Verifications
00:35:39
Speaker
um So that funding that they they did on the back of that idea, like ah was probably not you know, we didn't we didn't go anywhere. um Was the BGV for gig workers simpler? Less checkpoints in it? Yeah, well, there were three things that we did. One is the address, one is ID id verification. So we said, hey, first part is, ah so we wanted to answer four questions, right? Does Ashokaririn exist? Is he the one doing the transaction? Has he committed fraud in the past? Is he likely to commit fraud in the future? Right? These are the four questions. And broadly, as IDFI, that's the four questions we answered and today continue to answer.
00:36:18
Speaker
First one is, does Ashokararian exist? If he has an Aadhaar card, he exists. Or if he has an Aadhaar number and I can verify that Aadhaar number, he exists. um Second is, is he the one doing the transaction? So for for that, you need to know that Akshay is not pretending to be Ashok. So you do face match, you do some liveness detection, etc. This is the tech we built for the gig economy, ah workfor saying that, okay, first step is, let's take a scan your card.
00:36:44
Speaker
We OCR it, we check against the database, we then do a match against the face against the ID card that establishes that Ashok Haririn is the guy doing the transaction and that he exists. And we were the first ones to launch that, by the way. These were APIs that eventually we made it into APIs, but it was part of our platform.
00:37:05
Speaker
The first iteration of a convolutional neural net that we built in house. This is in 2015 that we
00:37:15
Speaker
building that, which is which then powered the ah financial services market, let's say, fast forward two years. In 2017, RBI regulations opened up. The same tech was used there. The third one was, has he committed ah third rulers has a committed fraud fraud in the past, um which where you do a court check, a criminal record verification. which And this is done online? Yeah.
00:37:38
Speaker
um We have mined every single criminal record in the country now, about 330 million criminal records. it's By an act of a government, it's public data. But these are in local languages, Telugu, Kannada, Tamil. It could be handwritten. So this had to be converted to a machine readable format. Again, we built our own tech for that. Again, very early... um sort of of tech around around and natural language processing um stuff, which ah this is a pre-ChatGPT, right? like So ah pre-transform models as well before 2018. So we built it old school way, which is all built in-house. We probably had a small team of our three AI guys when AI was not even a very big word. it's all 2015-16.
00:38:25
Speaker
so all two thousand fifteen sixteen ah But we built it on the back of the fact that 4000 Uber drivers would show up and we needed to verify them on the spot. They would show up physically to ah to a warehouse. Now, if I had human beings doing that, doing this collection of documents, etc., just slow down the process. You can't possibly onboard 4000 people a day the way we were doing it in the past. So that's sort of why we built it.
00:38:50
Speaker
And as we as we built that that, that was a first pillar of like real tech. um The fourth question, is he likely to do fraud? Yeah, is he likely to do fraud in the future? we didn't solve that yet at that moment. that That's probably a much later problem, which is where today's ID5 sits. But we there is a deterrent. um If you do an address verification, for example,
00:39:15
Speaker
ah that if you if you know the person lives at a place and you know his permanent address, he's less likely to commit fraud. Largely because he there's a society, you know where to find him number one and he doesn't want to to let his parents know that he did, he did Kant, if you may. so let's so So, yeah, so that's sort of,
00:39:36
Speaker
ah So address, so as on the blue collar BGV, it was ID verification check for the past listing and address check. That's all we did. And address, we had that 3000 people staff that we had built um in the first three years, they were able to handle that volume for us. So the middle part, which is the code check part, we ended up Actually buying the company, we used to take the service off. They had started mining that data and we bought that same company in 2022 eventually. Okay. yeah that's one of us You were their biggest customer.
00:40:17
Speaker
I was their biggest customer. Okay. like But it was called Crime Check and we bought them. but But we started taking their services because we didn't have that ah data and they had built expertise in that.
00:40:31
Speaker
Yeah. So that's sort of what we did, the three things, which scaled quite well, right? Because 4,000 people a day, there was a moment when, and within a year from 3.5 crores, had gone to 11 crores.
00:40:46
Speaker
Wow. Okay. So this is actually the true PMF moment, right? I think they say that like 3x, 3x, 3x, and then 2x, something like that is a formula, right? So so we went from, three no, I think we went from 3 to 8, 8 to 11, but anyway, 3 to 8, think, was the number. And and at that point, ah we started to feel that, okay, now we are seeing some early shoots of ah profitability, you know, now, and and remember I was telling you, talking to you about incremental, you know, um compounding, this is this is what incremental compounding is, right? You start that business, you you know, the basic idea is the same, but when you hit that moment where you need to achieve scale, you build that additional bit of tech, which then gives you the next scale, right?
00:41:33
Speaker
And then you again do that do that additional bit, like, so you don't You don't build for what IDFI is today. You actually build for what is needed at that moment and then and then slowly compound it over a period of time. um So that's kind of what what we kept doing, right? like and and And we were first in every one of those moments ah saying, okay, we are seeing what what the world is looking like and hey, can I can i build that extra bit of tech for that or extra bit of expertise for that?
00:42:04
Speaker
ah How did the move from employment or pre-employment verification to customer verification happen? I'm assuming when you talk of working with banks, you're talking of KYC?
00:42:15
Speaker
Yes. Yes. So, so so let's say 2015, did this. And RBI then came out with the regulation for digital So you could digitally verify people, um the first attempt at it, which is which is when the Aadhaar verification, if you remember, there was a KUA license that was given. um And we were a sub KUA. We started on in that path saying, OK, Aadhaar verification. we What's this KUA license?
00:42:42
Speaker
It was called, um I forget what the KUA stands for, um but essentially it's, UIDA is ah verification, like, so you would get a license to be able to use the yeah India stack to verify people.
00:42:56
Speaker
Okay, you can take the database and generate an OTP. That was the first big move. This happened in maybe 2015, 16, something around somewhere there.
00:43:10
Speaker
And we were a sub-KUA. So we were allowed we were allowed a license. I mean, we didn't get a license per se, but we were a sub-KUA. So we had we we were authorized by a license authority to be able to do this.
00:43:24
Speaker
And so that that again um improved the efficiencies, right? You could do you could do that verification very quickly. Incidentally, ah just a year after that, are we you know the Supreme Court squashed the Aadhaar verification thing for third parties and only REs, like a bank, could directly authenticate using using the Aadhaar stack because some there was this There was this court case that I forget who put it out also, but anyway. Something like about privacy, I guess.
00:43:59
Speaker
Yeah, something about privacy, which was actually not well researched, but anyway, doesn't matter. So they they they had put a stop to verification, which ah which actually killed a lot of my competition.
00:44:11
Speaker
And we purely survived that moment because we had the BGV business. Everybody else was just doing the doing the verification part ah actually died. like you know So you you you kind of get that scale and then suddenly your're you're, ah because of the Supreme Court verdict, it just like, it's hovers at that 11, 12 crore mark for the next three years in terms of revenue for us. And a lot of my competition like just practically died overnight.
00:44:39
Speaker
So ah banks were ah comfortable outsourcing KYC to startups? Yeah, I mean, um more than banks, it was fintechs which started with it first.
00:44:50
Speaker
ah The banks were banks are still doing it physically physically for a very long time. ah The first ones to actually... How did you break into this market? Like who were your first customers? and It was ah one of the crypto guys, cryptocurrency guys um who...
00:45:08
Speaker
who needed to do the verification because the government was coming down hard on crypto and they wanted to they wanted to make sure that terrorist funding and all of that. So they went with us and then maybe Ola Money because that's when we started to see scale when Ola Money went with us and we bombed that particular ah implementation, bombed it because suddenly the scale went from ah fuel a thousand verifications a day to suddenly like, you know, um ah you know maybe 20,000, 30,000 verifications a day when the wallet side picks up, right?
00:45:46
Speaker
And we're just not ready for scale. We had not built the platform for scale. We were built it we we had built it for literally like sub you know that that volume that we were doing. And ah oh like Ola was pissed off.
00:46:00
Speaker
um Obviously, they decided not to ah use us after that. oh um And it I think in ah in a span of about two weeks, we rewrote the full code.
00:46:12
Speaker
and end to end from what was on Ruby on Rails. We just took that thing and rewrote it in Elixir, ah which then I think they're one of the few Elixir companies in the country today um because Ruby was just was not scaling ah for us. and um And when we wrote it rewrote it in Elixir, we started to be able to handle that scale. And then we again, like went back to Ola and we kind of won them back. But but yeah, you know it its was pretty embarrassing when, when when when again, ah the whole idea of incremental sort of compounding is that, you know, you don't, some of these moments, ah the yeah how do you react to them matters. We we could have we could have reacted it reacted to it saying, hey, you know,
00:46:58
Speaker
how How can you rewrite anything in in two weeks? But we said, hey, there is no other way this is going to ah this's going to work. So we we took the whole team and we just told them, hey, if we need to rewrite this. We need to rewrite this in Elixir because Elixir can handle that that level of scale. And and they loved doing that. like I mean, I think for the first time, the engineering team really came together and just said, hey, this is a moment. They they actually canceled their Diwali vacation.
00:47:26
Speaker
um and And we all sat, like for a span of about two weeks, ah we were working 14, 15 hours a day just to get that thing to work in, in ah in a to scale the scale the platform. The moment we hit that, I think we start we were able to then handle about, you know probably probably about ah where where we were handling maybe two requests per second. We were now able to push that up to about 100, 200 requests per second. ah just okay just in that span of two weeks.
00:47:56
Speaker
ah And that allowed us to like then be able to handle the scale that wallets needed. ah Today, IDFI handles, I mean, we we have auto scalers now, so we don't even bother. um Even even at ah at peak, we are able to handle 100,000 requests second.
00:48:10
Speaker
a second which is ah which is Visa MasterCard scale. But ah that is we have we have handled that much, but we can it it's kind of unlimited at this point because because of our auto scalers that we run. But imagine like you know in those days, a lot of this was built piecemeal.
00:48:30
Speaker
ah For the wallet verification, what
Understanding KYC and Document Verification
00:48:33
Speaker
were you doing? What were the elements? Same same things. ID, you you take the card, OCR it, you figure out. Nowadays, we have also added tampering detection so we can detect tampering in any document for that matter. But then in those days, we didn't bother about that. We did then verify against source, do a face match, which we built um from scratch.
00:48:54
Speaker
um We probably, again, maybe the first ones in India to build a face match model. ah and And that's it. That was that was the KYC. Like I would take a photo of my driving license. Yeah, like you have we done your payday in KYC.
00:49:10
Speaker
That's us in the back. Long time back. I don't remember. yeah that's Take a photo of ID card, take a selfie. That's it. Okay. so That's from user experience perspective. And within about two seconds, you're verified.
00:49:25
Speaker
you You would, for example, but validate whether this is a real driving license, that the DL number or the pad number. correct Which would read the document, check against the database, ah do a face match between the ID card and the selfie.
00:49:39
Speaker
and ah And that's it at that point. Then in 2016, one more question. yeah A lot of identities which are state issued, how do you collate and aggregate that data? Like say driver's license. A lot of that is public data. So like for example, each state has its own database. Like so driver's license has multiple databases that you can check. By the way, all of this is is by act of government legal stuff to do.
00:50:07
Speaker
um Okay. Right, like these are ah aren not legal stuff to do, sorry. Act of government, it's public data, which can be verified, right? Again, just like Adhaar, like Pan. Isn't there a privacy concern that you're saying that I can go and look up anybody's driving license and therefore see the name and whatever data is there on the license. No, no, no. You can only check against the name. Like you'll get a yes or no. You won't get like, you won't get.
00:50:33
Speaker
and worse Okay. Okay. Okay. Got it. Got it. okay it's hard like You can't check, go check randomly. Like say, you can't check.
00:50:45
Speaker
what is Ashokharia and driver license that you're not going to get. Yeah. Okay. Okay. Unless you know there are competition of ours who who end up doing that, but that's all illegal. that That's they have they have scraped it or something like that, which is not allowed.
00:50:59
Speaker
The wallet business is part of the KYC, right? You were doing KYC. Our wallet business. Yes, it killed the KYC business because Aadhaar verification was no longer required. But fortunately for us, what we were doing is by that time, because of the, um because of Uber's use case, et cetera, we were not just verifying Aadhaar cards. We were also verifying driver driver's license. We were also verifying PAN cards. So when Aadhaar dropped, nobody could do anything, right? So But guess what? Only IDFI had the deal verification and plan card verification because we were doing it for the ah for the BGV business. and the And the fintech said, hey, this is allowed actually as a KYC process because RBI had that regulation.
00:51:40
Speaker
So we automatically got that business. so So almost like serendipitous because the guys who completely reli relied on the other stack actually failed miserably. So we kind of had a larger remit because, again, because of the BGB business, this one survived.
00:51:58
Speaker
It's fascinating how you have constantly been manufacturing luck. Yes. Like this is a case study of manufacturing luck, you know, being at the right place, right time with the right product. I think what we, that's what I would talk about incremental compounding, right? You just, you just, you continuously build and you build, in that process of building, you discover new use cases, right? Number one. Number two, you survive long enough, luck will happen at some point, right? Like, so, for luck, you need to, increase the surface area. Yeah, the surface area is longer. So, when RBI came up with the regulation, we were there. That's it. Yeah.
00:52:38
Speaker
Yeah. Yeah. Right. Okay. and and so and And our past helped us. ah So we we are today, for example, if you look at my competition in the financial services, none of them actually do BGV.
00:52:52
Speaker
and And BGV actually taught us a lot about ah risk and fraud, lot more than what my counterparts in the in the financial services side know. So based on that, we have also built tech, which is so different and so you um so unique because we are able to catch something which others just are not able to catch.
00:53:12
Speaker
That's also part of our part of our ah sort of offering today and I think differentiation as well. How long did this ban last? The the Supreme Court? Actually, it's still there. As a third party, I can't do Aadhaar verification. Only a bank or ah or only the requesting entity should can access Aadhaar today.
00:53:32
Speaker
Third party platforms like ours can't. Okay. But you are still able to play a part by facilitating this? Yeah, so what we do, um ah ah fast forward today, we have three platforms. We call it onboarding stack, risk and fraud stack, and the privacy and data governance stack. In the onboarding stack, um we we are a platform. like So we have the full onboarding experience that you can design. So taking those two photos, or if you want to fill a form, all of that can be done on that platform. Not only not only photos of ID card, you can do your...
00:54:04
Speaker
rental agreements, whatever you want, right? And it OCRs, it populates, it stores it in ah in a database, etc. in that In that workflow, you can drop your own API, Aadhaar API and say, here's my credentials or our bank's credentials. so I'm not doing it on behalf of IDFI.
00:54:21
Speaker
I'm doing it on but in behalf of my customer and the customer's credentials are used to verify. So it's ah it's for a customer, if it's like his own platform. For us, we are orchestrating a journey.
00:54:33
Speaker
and okay some of them Some of the verifications are ours, some of it can be third party verification, some of it can be my competitors own APIs as well, I don't care. and Okay. Moved away from what I would call just a verification business to an onboarding business.
00:54:47
Speaker
And that's sort of what the first stack of what we do today. So therefore, I don't care if if if India stack, suddenly in government adds more DPI. It doesn't matter to us because those DPI's you cant we can just we can just enable on our platform.
00:55:03
Speaker
What is a DPI? um Digital public infrastructure.
00:55:09
Speaker
like ah What do you mean? is it Public infrastructure is Aadhaar, UPI. okay Let's say but pan verification becomes... um what wites Actually, government does that ah have a stack for pan verification. For me, it doesn't matter because you can just drop that API on my platform and you can consume it.
00:55:25
Speaker
The way you want. So, a couple of years you said you were at 11 crore-ish. 11 crore-ish. We went, were at 11 crore-ish till about 2018 for two, three years. Again, the, what is that called? The, eighteen um for two three years um again the the the what is that call the Gig economy side didn't scale beyond that.
00:55:47
Speaker
And then 2018, once the RBI regulation, et cetera, picked up properly, ah ah our whole wallets, banks, et cetera, started to use the verification stack.
00:55:59
Speaker
ah that we had built around. and And so we were doing OCR, we were doing face match, we were doing, then we added something called liveness. We were the first ones in the world. ah In fact, whoever single photo liveness model, if you've done, if you've done your Zeroda KYC, they'll tell you to hold us this thing and recite a bunch of letters, right? That's called liveness, that's used for liveness detection. But our liveness, you don't need to do any talking, you just click us click a pick.
00:56:24
Speaker
And with with with a single photo, I'm able to tell if if it's a live photo or a photo of a photo or a deep fake. And we did this in 2018. We are the only ones in the world even today who's a certified liveness detection model with a single photo model.
00:56:39
Speaker
ah ah There are only 20 companies which have been certified globally. And we are only ones with with a single photo model even today. So if I generate a photo from Sora, you will be able to tell that this is ah well not a... well Yes, yes. But that is not the way this works. Basically, ah think of it as you would necessarily take a photo with a camera, right? and right As long as you're taking it with camera, you you then are able to tell if it's a deep fake or... or or if it has been kid or or it is a screen or it's a paper ah with a single photo model. Though ah you're right, we should be able to tell whether whether it's deepfake. But we are you don't need to worry about the deepfake on the web because that's relevant to us.
00:57:22
Speaker
yeah yeah yeah got it Specifically solving is in the journey, when I take a selfie, is it myself? Got it. Okay. Right. So you have camera control. Okay, understand. And we built, that is that is an AI model that we built in 2017, in fact.
00:57:37
Speaker
Again, for the gig ah gig worker use case, which then ended up getting used and in the in the financial services side. That's when the scale started to happen. And that's that's bringing you, like for us, at that point, we had now launched about 100 odd APIs of different types, from ah from liveness APIs to face match to other verifications to pan verification to OCR ocr kind of stuff. And we were the most extensive API sort of um house in the in the country. We had still not built our platforms at that time. This is just API, right?
00:58:15
Speaker
and And again, huge volumes. We've now moved from about 11 in 2018 to about 22 in 2019. So big jump.
00:58:26
Speaker
And then get voila, you have COVID. And we again, our revenue from 1.5 crores, 2 crores a month has now dropped to 30 lakhs a month.
00:58:42
Speaker
disaster. um yeah We are almost like we are about 250 people um we were just about scaling up. We are starting to see that trend line go up saying, okay, now verifications are happening. Fraud has started increasing because UPI had just started. And the moment UPI started, fraud started to increase, right? Because now scams were easy to execute because you don't need to be the person. You can do this over a phone, which wasn't the case earlier.
00:59:08
Speaker
um so So we were thinking that, hey, you know what, we finally have hit the hit that spot where it's going to grow. and And boom, COVID, And we are the, you know, March we had to, you know, obviously the company, we we went into remote work mode.
00:59:26
Speaker
And at that time, gone to like 30 lakhs a month. I'm guessing your pre-employment business would have come to a halt almost. Everything went to zero. Because nobody was employing. Everything went to zero, right? um But KYC should have still continued. No, because banks paused everything.
00:59:43
Speaker
Okay. Right. um but But what happened is in 2019, there was this video KYC regulation that came out that the only way to authenticate an individual digitally, they had to do video KYC.
00:59:57
Speaker
And we had built a video KYC tech ah on the back. Okay. And this is what happened. ah You couldn't onboard a customer ah physically. So everything moved digital and everything moved to video.
01:00:13
Speaker
And ah so first three months, we we go to 30 lakhs and we think we're dead. ah right We had, again, maybe two, three months of money left.
01:00:24
Speaker
ah we had a We took a call that we don't want to fire anybody, because including my security guard, who's not under my payroll, because we just felt that it was just a horrible time to do that. So we all picked and we we went to the company and said, hey, guys, we don't want to fire.
01:00:39
Speaker
What should we do? And everybody said, um ah that cut our salaries down. So we all massive pickets, senior leadership went down to 20% of our salaries, ah middle management went down to 50% of the salary and everything. And we made sure that we could last, instead three months, we could last six, seven months.
01:00:59
Speaker
Without firing anyone and and and stupidity in my head, I thought that COVID will be going for 2-3 weeks. It's not going to last for 2 years as ah as it ended.
01:01:13
Speaker
But ah after about 2-3 months, as the banks, as Video KYC was set to pick up and we had Video KYC Tech. Guess what? That started to pick up. even though ah Even though India went into complete lockdown, the only way you could digitally onboard customers is through what IDFI had built.
01:01:33
Speaker
Again, it's serendipitously tool to do that one. yeah Manufacturing. like yeah So, video KYC picked up. um We were the first ones to launch. we were abi We went to RBI and presented along with the bank. um And today we are one of the largest VideoKVC players. We have the full stack and we had to be able to onboard people in remote areas. So this video call had to function at 60 kbps. In fact, we are better than most of the zooms of the world in terms of being able to handle that kind of low bandwidth. um
01:02:08
Speaker
And we we built some spectacular tech there. and And again, ah that that that kind of gave us the second big big jump. But there was one additional thing that we did because Amazon was our customer and Amazon came to us and said, hey, our warehouses, we can't get people to come in. And Amazon was one of those critical services, if you may.
01:02:27
Speaker
um And and it they said, hey, we are we are a critical service. ah We are not able to bring people to the warehouse. So, you know, we we I don't know how to. We were just talking to them and they said, we don't know how to onboard. I said, why don't we use video to onboard your service?
01:02:44
Speaker
onboard ya delivery This is the the delivery, the last mile logistics guys. ah so the writers and Gig economy workforce, where Amazon was one of our customers or is certain of our customers even today.
01:02:56
Speaker
we We then said, hey, ah why don't we now digitize that and make that ah a make that a remote onboarding ah sort of reality for them.
01:03:07
Speaker
And that was on the back of video. And overnight, we instead of earlier, before that, I think Amazon had about 30 warehouses that people would show up and r tech would get used in in that onboarding experience.
01:03:18
Speaker
ah Suddenly moved to remote and we were able to launch 220 cities overnight, ah which couldn't have been done without video. So again, gig economy also picked up. So everything picked up in that in that COVID time for us because of Video KYC and the onboarding platform, we we had started to build at that time.
01:03:36
Speaker
And we built it on the back of gig economy again. So we were piggy packing on our old use case to build oh use cases for the oh for the but what i would call KYC business.
01:03:50
Speaker
And that's 2019. Then went from about 22 to about, 2020, we went 22, 21, went 37. 22, we went to about 57. so about ah sorry twenty two thousand and twenty we went we were at twenty two twenty one we went to thirty seven ah twenty two we went to about fifty seven 23, we hit 118, 24, 144, approximately there.
01:04:13
Speaker
um And ah yeah, like, I mean, it's it's just been a wild ride. ah How much do you estimate current year will be? It will probably be close to about 300. Wow. Phenomenal. You're doubling each year consistently. And you're thinking like, you know, if you think about it, 20, we were 19. We were
01:04:34
Speaker
2020, we were about 20 crores. 2019, we were at 11. So you're looking at like most of my growth has come in the last six years rather than the first eight years.
01:04:45
Speaker
Yeah, yeah, yeah, yeah. That's the power of compounding. Yeah, I get it now. Amazing. Right. So now fast forward to what we are today. We have three platforms. We have the onboarding stack. This is now used across pretty much any bank that you can think of probably uses us in some fashion or the other, either on the video KYC side, which is part.
01:05:04
Speaker
So onboarding can happen either as a self journey. or an assisted journey over a video call. right so So that's sort of the stack that have built on the onboarding side with all our APIs built in and inside that thing. So you can kind of consume anything you want ah on the verification side.
01:05:19
Speaker
The second stack we built is the risk and fraud stack. um Again, some of the tech that we have there is stuff that we built in the BGV side, which now ended up becoming the risk stack for us. And now we banks are starting to use, like the criminal record check, to check if somebody has a criminal case against them, um because that's ah that's a good indication for for fraud, et cetera. Second, um then we started to look at data and we said, hey, this data is telling us a bunch of things. So we started, now now we're getting to predictive side, which is the fourth question I wanted to answer is how do we how do we predict that this person might be a fraudster? So for example, If a person is in Daravi but taking a 10 crore loan, that should raise a concern. right so So there is address, there is ah there is data points, there are about 20,000 data points that we look at in order to figure out if there is possible fraud or possible risk in in a profile. um now being used x This is largely a lending use case? It's lending, it gets used in ah ah due diligence. um ah for For example, ah about 100% of the IPOs go through our platform for enterprise verification. Or not a verification, for due diligence, where we are able to tell, okay, but there is risk in this in this company.
01:06:32
Speaker
um And it's it's a SEBI regulation, by the way, so almost everybody goes to us.
Enterprise Risk Assessment by iDeFi
01:06:38
Speaker
at So what we do is we do individual verification, we do and individual onboarding, enterprise onboarding risk and asset onboarding and risk.
01:06:48
Speaker
So in the enterprise side, it gets used not only at um you know at at point of lending, but also at the point of like, you know of insurance, ah it could be used at at the point of IPOs, it could be used at the point of where where like an entity like Amazon is is it was bringing in third parties, they want to make sure that they're not risky, et cetera, because all this is...
01:07:12
Speaker
So the enterprise use cases, like, do ah does this company have a squeaky clean compliance track record? That's right. No code cases. code cases, round tripping. We can catch round tripping. um In fact, we we did this post on LinkedIn as well. Like, if if if people had just checked with us before giving Gensol their loan, ah we would have yeah you would have found out that the that was a fraud, not... fraud founders as well as company. you know There were signals of fraud in 2014 itself. So on the backup, we actually did a massive post on LinkedIn and we got a lot of customers because of that one post because they were like, oh, wow, well you you're able to find so much more. Because what happens with entity is oh you can start and any end number of companies. right So if you want to show high amount of revenue, you can start round tripping between your related parties.
01:08:02
Speaker
ah but So round tripping for people who don't understand is when you are billing your ah different entity and therefore creating some revenue. So Ashokaririn has company number one and company number two.
01:08:16
Speaker
Company number one is company number two's customer and company number two is company one number one's customer. So you can actually... sort of increase your revenue or kind of ah and obviously it's not as simple as this. It it could be multiple rounds so that it's not as easy as as just two entities. It could be this all the Havala transactions, etc. and I remember some article about this builder.ai and share chat something. Yeah, we we yeah we we found that as well. like We posted that. Oh, okay. I don't know. Our posts only on this. ah okay so so So we can catch round tripping. We can catch um whether they have court cases in the in the past. We are able to look at the risk from a more deeper perspective. For example, ah what is the impact of GST on specific industries and therefore are
01:09:08
Speaker
track track risk ah holistically. So all of this has now gone very deep, um not only at at at at at the enterprise level, but also at the individual level. For example, if um if your individual is um you know he's getting a salary in his bank account from, let's say, think and learn,
01:09:25
Speaker
ah that should be high risk. Why is it high risk? Because think and learn is by juice. How do I know think and learn is high risk is because by juice has not paid PF in the last ah three years and that's public data. So if I can, if I can go that salary came from here and then I can start making those connections to, data ah then I can start to figure out oh there's something wrong or or in in this entity um so we use a lot of that uh today to in order to figure out if if there's high risk on an entity or in on an individual we also because of our BGP business we also built a massive database of fraud companies who are in the business of giving fake certificates you you ran a recruitment business you know there's this massive industry of fake certificates fake uh fake degree certificates etc so we know yeah
01:10:09
Speaker
how this happens. So we have now a database of like 150,000 companies which are in the business of giving fake certificates, fake employment certificates. And and and that then we use as part of this risk taxing that, hey, this person is getting the salary from this company, therefore must be um must be possible fraud. And let me let me give you a little bit of an example. This might be an interesting one because for your audience as well as I'm sure you know this because of the recruitment business that you ran.
01:10:35
Speaker
What these companies do is Like Ashokaririn, let's say I want a job and because I don't have a job, I would go and go to this company, tell them, hey, ah can you give me a fake certificate? There are three ways that they do this. Either either they give you only a fake certificate and that's cheap.
01:10:51
Speaker
But if you want with salary credits to your bank account and they are ah they are legit MCA registered businesses who pay PF as well. So you'd pay them 10 lakhs. And for the next eight months, the guy is going to give you 80,000 to your bank account.
01:11:04
Speaker
or 70,000, whatever that number is, and ah APF, etc. Keep 2 lakhs to himself. So he he profits 2 lakhs. But now you have a record of a salary credited to your bank account and payslips which correspond to it.
01:11:19
Speaker
now But the company is a fake company. And there there are 100,000 these in India. right And now this number has now been expanding even more before it was only happening for employment. Now this is same guys are now getting used for loans, which then the people default on. So this sort of ah business now is is quite prevalent um in the loan use cases, which is also one of the reasons NPAs have gone up because your traditional models just don't work.
01:11:48
Speaker
ah Because now they have credit history, they're they're building synthetic identity, they're building synthetic credit history. ah But this person actually doesn't even have a job, right? So fraud all of a sudden over the last year, or two years has shot up like crazy because ah because of digitization as well as um as a virtual virtual transactions, as well as what I would call now fraud at scale, which is like people have scams as as a process rather than other than a one-off earlier.
01:12:21
Speaker
So that is the second platform. ah Right. ah Typically, banks have like a underwriting engine. Yeah. um I guess a lot of these underwriting engines are procured from vendors who provide the whole stack. ah so So where does your yeah so b API go? We would fit in between sometimes or sometimes we might be the underwriting engine as well.
01:12:44
Speaker
ah depend on you You provide a full underwriting engine also? I mean, they might use us in some fashion. Like, I mean, we don't replace. We are not a competition to... Like banks which are building it in-house would... Yeah. we would They would use our engine a little bit, but we don't replace our underwriting engine.
01:12:58
Speaker
What we are providing is signals. What we're providing is is the data science behind all of this, right? Like, for example, code data, nobody has mined. So, like, you can't do anything. You have to use me. um So so there are there are these pieces of stuff that we have built over ah over the last 14 years now, which are compounding at this point um for us because ah individually they were only so so valuable, but then collectively they become very powerful. ah right like So those APIs that we were selling earlier individually were worth maybe few paiseis.
01:13:32
Speaker
But when i when I put it together and start to put data science behind it, when I put LLMs behind it, um they start to add like significantly more value. And that's sort of what we are building at this point.
01:13:44
Speaker
Okay. So this is the second, the the risk business, the risk bucket. The third one is our data privacy ah product. um ah In 2018, when the ah when and um the ah the DPDP bill, at that time it was a bill, it was not an act yet, had just been... I've been...
01:14:04
Speaker
when I had been tabled, I started to get involved in in giving you recommendations on what privacy should be. And that came from a bit of a personal pain as well as personal angst around privacy. ah The idea was, I think one of my biggest problems is people have my data without my permission.
01:14:20
Speaker
and And for example, the builders who call you, I'm sure every but every one of us gets about three calls a day from Hajaar builders. They shouldn't have had my data. I don't know how they got it, in all of that. right so So I became a little sensitive to
Building a Privacy and Data Governance Product Suite
01:14:33
Speaker
it. And I felt really that, you know remember when I came back from india from the US, I felt I wanted to do some social service. so Now I had, this was my moment of giving back. And I said, privacy is my my calling card and and I wanted to contribute to it. So I started of contributing to to that act and over a period of time, worked very closely with privacy,
01:14:55
Speaker
with a lot of people around around thinking through what privacy might mean. And in that process, um we realized that there's a product we built. So in 2020, 2021, I built a lab stream inside ID5 where where ah there are 20 people, like we wanted to go back to the way we used to build things, like which is small team, can and we execute faster? So I built a lab stream in 2020, which then ended up taking this as a project to say, can we build a privacy stack, um ah privacy and data governance stack? So Privacy Act is not just about consent. It's actually about data. right
01:15:32
Speaker
Is your data private? is is Are you treating it correctly, et cetera? So that's sort of the thinking that we had. And then we built a suite of about eight products starting from consent and consent management to data governance and data ah and what I would call data asset inventorization and things. that I'll explain that in a bit. To breach management, in case a breach happens, what do you do?
01:15:55
Speaker
ah To third party risk, ah to um to then looking at it saying, okay, how do I do um ah impact assessments? How do I then look at my journeys and make sure that journeys are compliant? So there are about eight products in the privacy stack that we built from scratch.
01:16:11
Speaker
And again, we were the first ones to launch in India, ah the video KYC use case, ah because we saw this coming and not because necessarily we initially thought of it as a product, but we initially thought of it as ah as a problem to solve.
01:16:24
Speaker
And that problem then became a product. right Again, back to why did I start IDFI? It's literally about problem solving and and trying to take some interesting top ah exciting problems to kind of solve for.
01:16:36
Speaker
So in the privacy stack then we built that full stack and today we are probably the only ones implementing a very, very large bank privacy and data governance implementation. We announced that as well. This this product is called Privy. it's It's actually a platform about eight products which you can consume. Now let me let me explain what this means. Is that from a consent side,
01:16:56
Speaker
Think of it as you have a bunch of problems, right? One is, did I give consent? Can I revoke consent? If I revoke consent, where do I need to go delete my data, et cetera? What do I give consent for? This consent needs to be in all 22 languages according according to the law. Now, you can't have a lawyer look at each one each one of those 22 languages. There should be some automation to this. So we are able to then use an LLM to kind of translate to ah all 22 languages. And guess where that LLM got built? It got built from my code data product.
01:17:26
Speaker
because we had to build it. but use chat um We built our own LLM kind of thing, which which now gets used to do translations on on law.
01:17:38
Speaker
ah and Why couldn't you use chat GPT? yeah so So if you think about ah think about it, like one of the things, like for example, one is ChatGPD, it takes your data outside of outside of your environment, right? ah so But second thing is ChatGPD is not really trained on Indian languages, number one. Number two, there's very nuanced ah regulatory requirement here. Like for example, if you say pan-verification, I'm taking your ID card for pan-verification, you say that in English, of that can be converted to Tawa verification and in Hindi, right?
01:18:12
Speaker
and okay So that nuance needs to be trained and retrained on on our ah our models, which which is a model that we had built for code because we had already had done that for code.
01:18:26
Speaker
So again, leveraging what we had built, we built it for this translation, which is actually quite powerful because it automatically translates to all 22 languages if you create a notice. um Then there's a a consent lifecycle which which a user can go through saying, okay, i want to delete my consent. I want to have i have a right to review.
01:18:44
Speaker
and I need to know what I have given consent for and things like that. right That's your first part of the platform. The second part of it, which comes towards the end and I'll come to the middle afterwards, is it ah is around data. I have given you consent, but you have my data. But i when I say delete, do I even know where your data sits? Do I even know whether this data is stored properly. Do I even know if it has been anonymized, it has been pseudonymized to make sure that like Ashok's data is not sitting in some database and then a breach happens and then you are, because you you your fines are not on only on consent, it's actually on breaches. Breaches that happen on misused consent or on misused data or the data gets leaked.
01:19:25
Speaker
Now, so data, knowing where your data sits, not only where where structured data sits, unstructured data, for example, your ID card, your PAN card, your driver's license, et cetera, there needs to be a way to ah see that. And guess what? We had built all these APIs to do OCR.
01:19:41
Speaker
We plonked that into our data models. Now we are able to now structure unstructured data and give that ah data, um what we call our data compass, which is essentially a data discovery ah product, which is like an Informatica, but we are doing something more because we are now able to see ah JPEGs and and PDFs and things like that, which others, competition of ours cannot do.
01:20:04
Speaker
So that's all the data piece. And then so you can you can create your asset inventory. You can say, hey this data of Akshay, for example, is not encrypted. Encrypt this. Or it says, hey, this data, it it gives you lineage saying, where did I get this data from? I have Akshay's data, but where did it come from? Did it come from this journey or this journey? So that you know where the origin of the data, because when a customer asks, you need to be able to answer that question saying, hey, ah where did you get my data from? it should be able to say, hey, I got it from here.
01:20:33
Speaker
right so so So being able to explain it, explain where data came from, where is it sitting, how do you anonymize it, etc., is is our second stack we built. Again, based on a lot of AI and data models, based on our past experience.
01:20:48
Speaker
And then we get into what I would call breach management. In case a breach happens, even in case your data leaks, what do you do? um According to regulation, you need to answer to four people. You need to tell the principal that his data has been leaked, that is the customer. You need to tell the regulator.
01:21:03
Speaker
You need to tell certain and you need to tell DPB, which is a data protection board, which is the new regulator that's coming in for DPDP. Now all of them, you need to then be able to give them a remediation within 72 hours.
01:21:15
Speaker
you And for that, you need to now have breach management as a module. Because in case of reach, how do I how do i doubt inform ah people? so So the middle stack is all around around how do I do compliance and how do I keep my platform compliant? So that's around third party risk. if For example, a bank might have 19,000 vendors. like We are one of the vendors for a bank. I have PII data.
01:21:39
Speaker
In a bank, when we we are we are working at one of the banks, right we announced it as well, Access Bank. um is they have a crap load of vendors. A lot of them have PII data. Now you need to know which which person has PII data? How do I manage that ah manage that third party? Because if he leaks data, it's on me.
01:21:57
Speaker
So being able to handle that is that is is also part of the compliance compliance tool. um which is where we we are where we we do third-party risk and then we do impact assessments saying, okay, in this layer, layer what is the impact that I might have? what is what is Where is risk, ah for example?
01:22:13
Speaker
And finally, there's continuous compliance, which we built a product on where where it continuously monitors your ah your databases as well as your journeys and tells you we've gone out of compliance, for example, either because of product management and tinkered with it, or a regulation change, for example. Again, okay we are constantly reading regulation and and the platform is now saying, hey, you're out of compliance.
01:22:35
Speaker
So this is our third stack. right um So now if I go back, what we started with which is background verification to these three stack. and And by the way, background verification is an amalgamation of the three stacks. It's onboarding plus risk plus consent, built into one. So it's actually, if you think about it, it's ah it's BGV is one of our quintessential ah coming together of the three platforms.
01:23:00
Speaker
Yeah, yeah, yeah. This consent stack or the data privacy stack, I want to just simplify ah it a bit. So say, for example, if I apply for a credit card with, say, Bajaj Finance, so then when I'm in the act of applying, I'm sharing my data with Bajaj Finance and there are laws around how Bajaj Finance has to deal with that data. And that is what you help them to do, like in terms of... So three things. One is taking the consent and and making ah and creating the consent repository. right like So you need to know who all gave consent and it needs be mutable. That is, I gave consent today, Bajaj shouldn't be able to go and change that consent and say, i you gave me a consent for marketing even though I didn't. so uility got a three okay um Second is, ah so we are the consent repository. So we we say that Akshay has given consent and Akshay also has a right to review the consent he has given and change consent. okay So for example, today I might have given you
01:24:00
Speaker
um ah So consent for marketing on WhatsApp, but you started giving me shitload of WhatsApp messages. Now then I should be able to go back and say WhatsApp marketing stop. right and And immediately the company needs to now stop sending you your marketing messages. So that means your consent repository now needs to connect to your downstream systems.
01:24:21
Speaker
right ah Your marketing system, your your CRM, etc. This is one part of the flow which the customer sees. On the backend, there's a bunch of data oriented stuff that you need to do. For example, where where all do I have Akshay's data? So today, for example, if I go to a bank or any fiduciary for that matter, I might have given consent for, I said, no, don't send me marketing material, but I've taken a data dump.
01:24:49
Speaker
or in an Excel and given that Excel to some person and that guy is sending marketing, right? How do I know where my data sits in my whole Akshay's data is sitting and say, yeah, all these places data is sitting. So we need to make sure that each of these have access controls in the right way.
01:25:06
Speaker
art method ah Does this person require data access? Because you want to minimize the amount of data access within your environment as well. So so being able to understand where data, the So where the data sits or how you treat that data is something that's the second platform.
01:25:23
Speaker
ah inside Inside that's what we do. Third thing is around compliance itself. So like, how do I know if my journeys are compliant? How do I know if there is suddenly there's a risk because there's a new third party ever onboarded, but that third party is misusing data, have they deleted data, things like that, which is all internal to ah to a bank. And that's where like lot of the compliance part lies. but ah for From a user perspective, ah you will see our consent upfront.
01:25:50
Speaker
From a fiduciary perspective, we do all the magic in the back end. Okay. ah I believe in the European GDPR law, there is a requirement that companies cannot keep data beyond X number of months and they have to delete it. and Is that there in India also? Same thing there in India.
01:26:07
Speaker
that's So when I'm applying for a credit card somewhere, I'm sharing my data, then that company is supposed to delete my data. um so so So there's RBI regulation, which says that you you keep the data for 10 years, I think, seven years or 10 years, depending on the regulation. um So they keep it until that time, but they need to delete afterwards.
Navigating Data Regulations and Industry Reactions
01:26:25
Speaker
Number one.
01:26:25
Speaker
Number two, you can only use it for the purpose that you have given consent for, which means if i'm not told if I've told you do not do cross sell to me, then you can't do prosper doing um But nobody follows this. The act has not been, the rules are not yet out.
01:26:42
Speaker
So the moment the rules are out, ah then ah within about two years, they all need to get compliant. The rules will be out any time. It's being vetted by lawyers right now. It'll come out any day.
01:27:00
Speaker
This would even apply to, say, retail. like say everything but anybody Anybody touching consumer data. like right It could be retail, it could be car OEMs because they have consumer data you might have bought from Tata. If you bought a Mahindra XUV or XUV 700, then your name might be there and they have data then it would apply to employees because they're they are also customers, that's PII data that you hold. So everywhere everybody comes comes comes under scrutiny. I think it'll take at least four or five years before
01:27:33
Speaker
before people get fully compliant. But I think the processes have already started. Are tech startups welcoming this or do they see this as a disruption to how they do business? I i mean, it's an easy way to market, right? like Yeah, I think I...
01:27:52
Speaker
See, there's a whole spectrum. Some guys hate it Some guys reluctantly say yes. Some guys say, hey, this is an amazing way for us to differentiate as a trustworthy platform.
01:28:04
Speaker
And I think in that spectrum, you'll find more people towards, I don't like it and I want to fight it than the guys who think of it as a trust building exercise. I believe the guys who take it as a trust building exercise will win.
01:28:18
Speaker
And I do think the banks, for example, are taking it very seriously ah with like ah probably ah some of the looser startups taking it ah not so and not in stride, but I would say that I think anything anybody ah RBI regulated is is taking it very, very seriously, number one. And things like banks like Axis have taken it to the other extreme saying we will be the trust the bank that you can trust because they are the first ones to go and go in and say we are going to implement this even before the rules are out.
01:28:56
Speaker
okay You have that spectrum. So you're again positioned for luck to strike. Like like the moment this becomes an act. It's already become an act. Now the rules will be out anytime.
01:29:07
Speaker
ah So yeah. what What does that mean? It's an act but rules are not out? So there's an act which says DPDP is a law already. um But there's rules of implementation which they have tabled. ah Okay. Which will come out ah in the next...
01:29:22
Speaker
hopefully next we by Hopefully by but winter session. if that Once that comes out. Which is like timelines, how much time you have to implement. Specifically saying what all you need to what are you what all you need to do in order to get compliant.
01:29:35
Speaker
Post that, the DPB will be created. And the whole post the rules coming out, I think the implementation process starts. um But yeah, I mean... um It's any time now. and yeah and And you're poised for that yeah surface area of luck. Amazing. Okay. These three stacks, what is their contribution to revenue? I think Privy is just new, relatively new. So it'll start now. But the other two are 50-50, I would say.
01:30:06
Speaker
just Okay. Who are you competing against like for the KYC stack? So if I look at the APIs, ah which is only about again again, BGB is one part of my business, API is another part of my business. The platforms are the third part of my
Growth through Strategic Acquisitions and AI Development
01:30:21
Speaker
business. business If I just look at APIs alone, it would be the likes of like, I guess, Perfios would be the largest guy out there. um But on the platform side, like Video KYC, for example, i am the largest. ah and And my competition might be somebody like WorkApps, for example, which is the second largest. right okay But I'm the largest by a long mile. But if you look at onboarding, ah it could be like somebody like an Adobe, which which is an onboarding stack. you know It's a very different competition. If I look at the IPO business, I'm 100% market share. So it like like there' is there is nobody else. On code data, we are we are the you know we have the largest database of code records. So we are at... So like very, and privacy, i'm I'm mostly competing with global players. There is there is no player in India who's of our size at this point ah or having any any implementation record. Like we are the only ones, but now we have seven customers who are already implementing or
01:31:24
Speaker
are going live at any any time now. um So so they I think we would be the largest there. So very different kind of, and BGV, it would be like the old first advantages in Northbridge. So I'm like kind of ah split between like three, four different types of ah competition. It's not the same guy in in across these three.
01:31:45
Speaker
How much did you spend for acquiring that quote? ah Yeah, I would not like to say that number. Okay. But but that's that's like, ah again, a very strategic acquisition. Yes. and and And, you know, because of that particular part of the business that ah from when we bought them to now, it's it's grown at least like 7x, 8x now.
01:32:08
Speaker
The revenue was just because of what we've been able to power with it. Yeah, yeah, yeah. Code case on an asset as well, right? Code case on an entity or code case could be on on an individual. um So you can actually then start to build risk signals. For example, there house which is which has which has litigation? Then you might not want to lend to that house, right? Even though, yeah, so there's lots of things we can do with the data that it has.
01:32:36
Speaker
ah you You said you built your own LLM or SLM? We built an SLM. Okay. In 2018, we did before the Chagibiti thing came out. um Because we the transformer model came out, if you if you remember in 2018 when Google published it, we read that and we built our own to solve both solve for language issues, language problem that we had, which then we used when we bought the code company, we used that to um to
01:33:07
Speaker
that language model to kind of uncover like some of the data that we wanted to uncover from that code. And then but then got used in in in the privacy world as well.
01:33:19
Speaker
So one use of AI is translation. So there were three AI stacks. One is what what I would call computer vision. Computer vision is old or what do you do with an image. You can OCRA it, you can do face match on it, you can detect tampering on it, you can do um you know that deep fake detection on it. you can like and You can do a bunch of things or you can... and On a video, for example, which is an image or or stream of frames, I can tell you but what if this person is saying or what like you can look at micro expressions and say, is he lying and things like that. um okay so So that you can do on image. So we do a lot of work on image.
01:33:55
Speaker
um Then we do natural language processing, which is around using LLMs or SLMs. So what we have is we have retrained ah LAMA 4 now um ah and and and then put our SLM on top to to do some of the things we want to do. um So that's something which we do. lama We do Vikuna as well as LAMA. So we we're all in-house, by the way. We're not wrapping on a chat GPT.
01:34:23
Speaker
right yeah and then And then the third one is data science, which is old school data science, regression models, et cetera, which is not necessarily what you would call AI or ML, but ah uses some ML models as well around around trying to figure out time series data. right like So if I know time series, then I should be able to tell what kind of sort of risk. I can put a little bit of ML on it. We can put a little bit of what I would call old school regression models, like
01:34:54
Speaker
linear regression or multivariate regression. What is the benefit of building ai tech in-house versus just building a wrapper? um There are two things. One is ah the price points in India. ah you can't You can't deliver at this price point if you don't build it yourself because it' just the costs are just prohibitive, number one. um so So that's... it's like But costs are constantly going down, right? like next year ah constantly It isn't really, but especially if you have to use ChatGPT for everything, the costs and that goes up significantly, right? Because it doesn't work for us.
Fostering Company Culture and Innovation
01:35:31
Speaker
Second, you're sending data out.
01:35:33
Speaker
out of your premise which yeah okay right third is Third is something else. But this sending data out is something you're uncomfortable with or it is regulatory not allowed?
01:35:44
Speaker
um We are uncomfortable as well as regulatory not allowed. Okay. right. Third problem which is which is that these guys can anytime shut you off.
01:35:56
Speaker
right What happens when you... like rappers, in my opinion, are um like and a lot... there are lots of startups now being built on Agentech and that's all wrapped around ChatGPT. For tomorrow ChatGPT shuts you off, that business is gone.
01:36:10
Speaker
ah Right? um and and it and and And we have seen time and again this happen. So I'm very paranoid about it saying... and and and it's not that hard to build. I mean, it is hard to build, but it's not...
01:36:23
Speaker
um It's not prohibitively hard to build specialized models. um It's very hard to build a large model, but if you're building specialized models, it's not very hard. And you can also get precision to be far better than a chat GPT. For example, if I'm specifically looking at RBI regulations and I know patterns on RBI regulation, I don't need million tokens in order to decipher what RBI is saying. And I can do it in a far more efficient manner by doing it you know just a 20 token model.
01:36:53
Speaker
Right? Hmm.
01:36:56
Speaker
so Okay, I want to kind of wrap up with a little bit on people management insights that you've gained over this decade and a half. but What's your headcount today? We are about 800 people.
01:37:08
Speaker
Wow. How do you successfully manage an organization with 800 people? What kind of mindset changes did you go through from managing 12 in those early days to managing 800 now? Yeah, I mean, there's one thing we did very early, I think, which did us a lot. I mean, it's a reason that we, I think we are as good as we are. is we we invested in culture on day one.
01:37:30
Speaker
um And for us, it was very clear that you know if if you ask anybody what was the best time they had in their life, nobody is ever going to say work. right Everybody's going to say college and things like that. And I kept wondering why does why does work have to be drab? Why does work have to be painful? Why can't work be like college? And the three things that mattered there was you are allowed to make mistakes, you're learning constantly, and you're making deep friendships.
01:37:58
Speaker
um and And the fear of mistake is what makes work very drab. And the way we thought about it is if you're not allowed to make mistakes, then we then we are not going to be able to push boundaries. Like people should be kind of open about making mistakes and not getting consequences because of a mistake. There might be consequences for laziness and things like that, but not for a mistake. So we so we we really double down on that, saying mistakes are forgiven at ID5. Please do more of them.
01:38:25
Speaker
um Mistakes as in like actual coding mistakes and things like that because we want people to push that boundary. Second is we want to invest in shitload of learning, right? Because we are we are bringing in young people, they should be able to come to a place where they can learn. This should be a seat of knowledge. So which is why pushing those incremental compounding that I was telling you about every time my my team is constantly learning new things.
01:38:47
Speaker
from Ruby to Elixir, now AI, right? And everybody's full stack. We don't have like, we have an AI team, but we don't have teams which are specifically back and front end, like you you deliver the whole end to end. So that kind of allows them to learn a lot more.
01:39:02
Speaker
And the third one was deep friendships, right? Like if you come to IDFI, most people will tell you their best friends are here. um right and and And that kind of creates, um so there we focus on three, four things. One is, why do you feel Indian? You feel Indian because of shared memories, shared stories, shared symbols, ah cricket team, you're rooting for them.
01:39:23
Speaker
um It is an artificial construct, if you may, the boundary. right Why are you considering to be Indian? Because i have a flag, I have an anthem, I know Gandhiji fought for us, I know ah religiously there's Buddha, Ram and all this stuff which we tell stories about.
01:39:40
Speaker
um And a lot of those are are things which we embed into the org saying that can i build um can I build culture, culture cannot be built if if you're not investing in those those symbols, those stories, those memories. And these memories become bigger than the... ah You know, everybody talks about it, even though they might have joined ah later compared to the original twelve.
01:40:01
Speaker
They talk about the apostles. So can we can we create this these memories which are... you know you for you For example, you probably... very fondly can think about, oh, what must have Gandhiji done? It's my our memory. it's not It's not people who lived around Gandhiji's time. it's It's India's memory. right And that's sort of what we wanted to build to kind of create that environment of culture. So we invested very deeply in that.
01:40:28
Speaker
Second thing we did is we said we're going to share a lot of wealth. um So ah we have one of the largest ESOP pools in probably in the country. ah ESOP pool is about 15%, which is unheard of in the market. We said there should be enough people who make money, um not just the founders who should be making money. And if you think of any of these IPOs and of these startups, not too many of them are sharing wealth. I don't understand this.
01:40:52
Speaker
Only the founder makes money or five people make money. It doesn't work. So for us, it was very important that we took as inspiration from the Narayan Mutis of the world rather than the ah the heroes of today, if you may. And that's sort of the second second big layer.
01:41:09
Speaker
Third one, which which I think also we we ended up investing a lot is we want to push decisions to the last mile. if your If management is taking all decisions, then we'll be slow.
01:41:21
Speaker
ah We can't do this innovation, we can't do this incremental compounding that we are talking thinking about if the last mile is not taking these decisions because I won't know enough information. I don't know what is going on a lot of times. right So can the last mile take more ownership? And we gave that as the third lever.
01:41:39
Speaker
And between these three, I think we created an environment where we've we find that people tend to not leave the company, number one, and feel that it is their company, not just Ashok's company.
Planning for IPO and International Expansion
01:41:54
Speaker
And to that effect, we also have now put in place the fact that you know just because I hold this position today, I don't need to hold this position tomorrow. This is not a...
01:42:06
Speaker
This is not Ashok, the founder's company. It is Ashok, the management team's company. Founder is ah is the wrong word like for us. Co-founders don't have these titles internally. We say this is run by this leadership team, which can change over a period of time because that's what we built on.
01:42:27
Speaker
The org needs to be built for 40 years, not for not for a short period of time. So that's philosophically what we're building. So all of us now think of this 70, 80 people that now run the org. um and And we want the next bit of leaders actually coming from within.
01:42:42
Speaker
a Long story, but hopefully it will be as of what... No, it's inspiring. Some of that inspiration came from how HUL runs. I'm really inspired by Unilever, right? Yeah, especially Hindustan Unilever as well.
01:42:55
Speaker
And that's sort of where I thought, you know, if we can build that kind of culture, then you really can build great products, right? So this is the value of doing an MBA, right? Being able to think like this, right? I think MBAs, we are not thought taught to think like this. like we We are taught, I think, largely to be Excel junkies, you know or or for that matter, like you know strategic competitive advantage. We don't think culture at all.
01:43:21
Speaker
um In fact, for me, the concept of culture actually came from the company that I worked for in the US, which had this and it was an engineering led org. And i from engineering, generally, if it is run by engineers, you you you intend to build, I think, more fundamental thinking ah models rather than just a puristic, ah it's a pure... um commercial model, right? MBA is a commercial sort of way of thinking.
01:43:49
Speaker
And I think what we have built here is probably more more holistic and and and and deeper than what, I mean, I don't think any of the MBAs, like Narayan Moti is not an MBA, right?
01:44:04
Speaker
Yeah, yeah, yeah. He built Nandan Nileke Nile, but think of what they have built. Yeah, true, true, true. I don't know if too many. ah And so that's why i kind of feel like the engineer in me is constantly tinkering and saying, hey, how do I how do i take this to the next level?
01:44:22
Speaker
What's the ultimate goal? An IPO? Yeah, I think we'll do IPO at some point for sure, sooner than later. But the ultimate goal is bigger than that, right? I want to be at the center of every sensitive transaction in the country or in in in in this region. So we now ah ah do operate in Philippines and Indonesia and MENA as well. We are the second largest player in Philippines.
01:44:44
Speaker
So we want to now ah own that and say, okay, in every sensitive transaction, can we be involved in assessing risk in onboarding or in in doing this privacy thing? um And that's what that's really the goal really, really right and and being able to authenticate that many number of people etc.
01:45:01
Speaker
And I look at it as a as more of a mission of enabling genuine people to get opportunities rather than catching f fraudsters. Like frauding catching fraudsters one part of it.
01:45:11
Speaker
The real thing is can can genuine people get more opportunities right and and if we can enable that I think we have we have we have achieved our goal. And and and i'm I'm kind of proud of the fact that maybe we have been pivotal in that, when in the in the gig economy workforce space or in in and people who today can get loans who couldn't because ah we have better models today.
01:45:36
Speaker
ah What percentage rate of your revenue is India? um We have about 15% is international. Okay. um and and And until a year and a half back, it was zero. Wow.
01:45:47
Speaker
Okay. Amazing. Amazing. Thank you so much for your time, Ashok. It was a real pleasure. Thank you so much, Akshay.