AI After Dark

0:00 The reason I'm excited about recording with you is because I think I've had the benefit of being, while I'm Mercury Fund, access to all these AI native CTOs, especially over the last year. And I

0:12 get to listen to you guys talk about how your thought processes are, how you develop teams, how you got to where you're at. And it's exciting for me, but I've had enough people that reached out

0:23 and are like, man, I want to invest in this, or I want to be part of this, or I want to build this, and don't have exposure to the people I do And so to the extent that I can create a repository,

0:32 basically, of those nuggets, I think I'm excited for. So thank you for having me. Thank you for thinking of me. Absolutely. Of course. So I want to, like, I was thinking of just, like,

0:42 telling your story a little bit, and then, like, as a background, but I know pre-recording you were giving me some little bits and pieces of it. Like, I think we need to start even earlier. I

0:51 think we need to start with, like, like elementary school. Give me, give me, bring me from elementary school to today. in your background, 'cause I think it's fascinating. I think there's one

1:00 thing that probably kind of sets the context is I changed seven schools growing up, right? So my dad was in construction. We moved from one destination to another. He had built hotels and resorts.

1:10 So never made a lot of friends growing up, right? 'Cause it was like one year and then you moved to the next school. So I think that set me up for being an introverted kid. Spent a lot of time

1:20 with myself thinking about things and how to solve problems, if you will, and be adaptable. And now you're in front of a microphone. Now I'm in front of a microphone Yeah, my dad was what me

1:28 growing up. He's like, I don't know what this kid's gonna do 'cause he's so shy and timid, right? So from there to here, it's been a quite a journey. And I was telling you, I think when we

1:37 started, I started coding in eighth grade. My first program was in C programming, right? We did like flow charts and very interesting stuff. I was very fascinated by it. I was really good in

1:48 STEM always, right? I think I sucked in history and those kinds of subjects. I never really liked them Computer science I was very drawn to.

1:59 Sorry, 'cause I'm the student that got a C, and C, it was very poetic. That's what drew you to it, Nissi. I think I love the structure, right? So there's like, if this do this, if something

2:14 else do something else, like there's very logical thinking, and probably I was missing that in life, right? And it kind of grounded me in more ways than one. Thinking back, that's what I think

2:25 it was Got it. And then I also loved that you can build kind of that zero to one action. It's like you build something out of nothing, right? You have this idea in your brain, you enter into the

2:36 computer, and then it builds it for you, right? So I think I always found that extremely fascinating. Could be very simple stuff. Like, I remember the first thing that I built was like a

2:46 calculator, right? Very simple, you have a calculator, but it was really awesome to see that live on a computer, right? from scratch, quote unquote, right? Not like I was writing assembly

2:56 code. I didn't really do any of that. It's very like Java or C, very abstracted out languages, but it was very fulfilling. My eyes lit up and I wanted to do that on repeat. Is the mindset like

3:08 you deconstructed the calculator on a software basis to reconstruct it? Or no, no, you learn the tools, the building blocks and put it together. It's yeah, like deconstructing the calculator,

3:17 right? Like, and then the same, I think MVP thinking, right? Like, okay, let's start with plus, minus, divide and multiply, right? So just start there, and then you can add brackets to

3:28 them or braces that are coming from indents and brackets, right? So you're kind of adding extra layers, and I think that MVP thinking started right there. It's like, let's start small, and then

3:38 keep adding to it. And I was telling you this before the recording began is I didn't have a computer growing up, right? So a lot of my coding was on with pen and paper, so I'd write it down.

3:49 write multiple versions of it because I was a little worried that maybe one version wouldn't compile, there'd be some errors. So kind of there'd be like two or three versions. And then I'd get one

3:58 hour at a computer lab to go and kind of type it in and then compile it. And then the worst part Alex was if it didn't compile, right? Some random syntax error, I'd 45 like spend minutes debugging

4:09 it. And the hour is ending, you have to kind of wait for another week to see it in action. You'd get like a ledger of your debug, your bugs, so you could go Yeah, we could print it and kind of

4:19 go back home and kind of see where it went wrong. But you have that hour, you want to like be very productive and efficient about it, right? So, but that being said, I think I became a better

4:29 programmer as a result of that, right? So I had to write really good code without compilers, right? And then go compile the code. Well, I think something, and I know you're a humble guy, but I

4:40 think something that you probably do better than most, especially me is like, let's say I've got a bunch of things, and I just call it a grocery list. When I go to the grocery store, I can't find

4:49 one item, even if it's the third item on my grocery list, like I will spend an inordinate amount of time trying to find that item before going to the next, which is like this task rabbit mentality

4:58 that destroys productivity. And so this learning process, though, forced you to appreciate the prioritization process of what matters and what doesn't. 100 is that I think what is important, what

5:09 is table stakes, and what is a nice to have, right? So kind of very much that thinking came in very early in my life, I'm very grateful, grateful for that. And I think I was able to then apply

5:19 that throughout my entrepreneurial journey, right? Kind of take that. In fact, my first startup was not a tech startup. I wanted to do a events management company, right? As an introvert. Yeah,

5:32 as an introvert, it's great. So I think we, I stumbled upon it because in India, I went to a private school, very grateful again for that. But we had like a very robust school parliamentary

5:44 system where students would elect. 150 students to lead there, lead them, right? So it's about 5, 000 students, big school. So it was Minister for Science and Technology, Minister for Sports,

5:58 blah, blah, blah, President, Prime Minister, trying to kind of get people to understand how democracy works, how you vote, and all of that kind of stuff. But so I became the Vice President and

6:06 then President and I was organizing events for my school. The last one being a Alua Twist play where we converted our school building into an amphitheater. Wow. And I got to direct that. So it was

6:19 awesome. That was like, that kind of experience you don't get as a 17, 16 year old, right? So it was your leadership style at 16, 17, 18. So if you should ask my friends this, I'll probably

6:29 tell you, I was a trickler, like I was very strict, right? Very much about the rules. Like I'm sure a lot of people hated me for it 'cause I gave them a lot of hard time. But yeah, I think it

6:40 was very much, probably emulating my father, right? That's the best, I mean, for me, That was the closest reference. I've evolved a lot since, right? I think I've developed a lot more empathy

6:50 since, but then it was like, you know, get stuff done, be angry with people who, if you're now falling in line, so that kind of stuff. So yeah, come a long way from that as well. But my first

7:00 company was going to be an events management company, is because I realized there was so much inefficiency in just that industry as well, right? And if somebody could just streamline, they didn't

7:10 even know how to use Excel spreadsheets to organize themselves. Everything was in their heads, right? So it was over, hey, here's a spreadsheet. All of this event is organized nicely in this

7:21 spreadsheet. Everyone knows what they need to do. It's very good. And then we can track, this is 2008, 2009, right? So India is still kind of catching up, it's especially like ground level

7:31 stuff, right? I'm sure companies were using all of these tools, but your regular like a chair's window, he doesn't care about a spreadsheet, right? But like it was easy to organize people and

7:41 kind of get them on the same page. So I was like, hey, maybe there's like, efficient way, then we can kind of save a lot of money and the upkeep is more, right? Like a row, low, if you will.

7:50 So that was the first company, again, didn't end up doing it. Tech just happened, but anyway, a site, a quicker site. When you say tech just happened, what does that mean? So

8:03 the other aspect is I was preparing for the IIT, which is the Indian Institute of Technology, the best institute in India

8:12 I spent about three or four years preparing for it. So this is ninth grade through 12th grade. So we'd go to school from eight to two pm. And then two to four was extracurricular activities, like

8:22 the president stuff. I was a school cricket team captain playing cricket. And then from four to nine pm. was these extra classes to prepare for the IIT. And that was four years every day, right?

8:34 Including Saturdays and Sundays, Sundays with extra class, right? Wow. So I barely spent time at home because it was either Bangalore traffic or these classes. So the whole plan was hey, we'll

8:46 go to IIT, you know, that's that's the best place to be in India Never thinking about the US My dad forced me to apply to University of Texas Antonio because he's like, yeah, just a backup habit,

8:58 right? Why UTSA? So my grandfather was a professor at UTSA

9:03 He wasn't he wasn't alive when I actually ended up going there, but my dad has a fondness fondness for UTSA And he's like, you know, just just have it as a backup. What do you lose? I literally

9:13 didn't prepare for the SAT I think I just showed up and it took the

9:18 SAT So IIT happens. I don't know how if you're familiar with how competitive it is to get into the idea No, but just but just within like the the training program. I can imagine. Yeah, so I got I

9:27 think 200, 000 people take the exam at the time and Probably more now and maybe four or five thousand people get it right so extremely competitive And I think I missed that range by four or five

9:40 questions kind of going back because there was negative scoring for any question you got wrong, so it kind of dings you for it. So it was four or five questions and I missed it. And I was like,

9:51 I'm done with this. Like, I'm not telling you the whole story, but actually, like, I think I took like 11 exams back then for different schools. I got into a bunch of them, like all the tier

10:01 two schools, but I was like, I have so unhappy and devastated. I was like, I need to leave. And that's probably the best decision in hindsight, right? I mean, UTSA, a lot of people were like,

10:11 yeah, it's a no name school. Sorry, but like it was back then, right? It's like, hey, you know, like, I don't know if this is the right decision. Like my family was weighing in in India,

10:19 they do that, right? Everyone weighs in and they're like, yeah, it's just the best thing for you. But in hindsight, probably the best, best decision. And that's how I got into tech, because

10:29 my first year in college, I go to the computer science class. After the second class, the computer science professor is like, dude, I don't think you need to be here like just take the challenge

10:38 exam just keep this because Yeah. And be my TA next year, right? That was the feedback. Yeah. And then the same professor gives me a job at the Department of Computer Science and he asked me a

10:49 question. He's like, We need a website. Can you build it? I don't know how to build a website. I was like, Yeah, I can build it. It might not work pretty, but it'll - No, I didn't even say

10:59 any of that. I said, Yeah, I'll build it. I'll build you the best website, right? Like the best website, I don't know how to do it. I didn't know how to do it, right? Had to go figure it out

11:07 'cause I'll never build a website Like I think writing Java programs is very different from building web applications and websites. I think people kind of looking at tech don't realize the difference.

11:17 So a lot of people who study computer science are not always equipped to build these web apps, these SaaS apps that came out. So they had to kind of go through extra training to be able to build it.

11:26 I didn't know how to do it. I was like, Yeah, I'll do it. And then went to figure it out. I think that became the motto of my life. You just say yes to things that come about, right? Well, I

11:35 don't think I don't, know if you, maybe you don't realize how rare that is because you are that, but bye. like having the arrogance to say, I can do this and the humility to actually go build it.

11:45 That combination doesn't exist in a lot of different corners. I'm grateful. I don't know. But I'm

11:53 curious, how do you think you developed that? Was this something innate? Do you feel like this was part of this process of trying to appreciate the logic behind building software? Where did that

12:08 combination of arrogance and humility come from?

12:12 I think I had a chip on my shoulder. It was like, I want to prove that I can do anything. Growing up, we grew up in a middle class, upper middle class family. We were always trying to strive to

12:23 be more. And one way to be more is make a lot of money. That's how you get influence in the Indian society, too. It's like, if you move well off, you have more in an easy scoreboard. Exactly,

12:34 right? So I think it came from a sense of like, hey, I want to do more. I want to be, I want to make a lot of money, right? I ended up doing a tech. That's a funny story, but I kind of want

12:46 to do more. And I think the part of me doing a tech is I still have rose color glasses. I saw this perfect world. I feel like we could use technology to solve real problems. And that's where I

12:55 think the combination of the arrogance and then the humility to kind of go do it, probably come from.

13:03 I don't know, Alex, I don't know if there's like a, sorry, I don't know if it's the best answer. No, no, yeah Yeah, to me, what that says is this is something that was likely instilled in you,

13:13 you know, from either a young age or just innate, and it's something you continue to refine. I mean, like, you know, I'm from Argentina, people are like, oh, Messi, he was born with it,

13:22 right? And I was like, yeah, but he also had to develop, you know. 100, 100. It's a lot of nurturing, right, versus nature. Because my nature, like I said, when I started, I was a timid,

13:32 shy child, right? So that was my nature. I think nurture is like looking at my grandparents, and how they went about their life and taking the best of their colleagues. My grandfather on one side

13:43 was IIT professor. He was a department chair in math, right? And the other guy was a national tenant. He was

13:50 a student. I just didn't get in. Imagine how he must have felt. And then on the other side was a national level tennis player. So that's the combination. But both of them had a lot of discipline

14:03 that probably took that and a really great work ethic, right? Like just kind of working, putting the hours and getting stuff done. I definitely got that from them, right? I saw them in action.

14:15 I was very lucky to see them in action 'cause my grandfather would take me to work and I'd just sit there with him, watching him work, right? And that leaves an impression on you as a child. Yeah,

14:24 yeah. So - Mine was the other side of that coin. My grandfather was anesthesiologist, chief head of staff or whatever the term is at a hospital. And when I would go to work with him, by all the

14:35 things I saw and veered far away from any medical profession. But it isn't amazing like the impact that those moments have on you, even at a young age. I don't even remember the moments, I just

14:47 remember how I felt. How it felt exactly, exactly how it felt and how you want to feel, right? There's some situations where you feel terrible and you want to feel different, right? And then

14:55 there's some situation where you cherish them and you want to feel that again and again. So I had both, I think - Well, great heritage. That's awesome. Yeah, yeah. So your UTSA - UTSA, UTSA,

15:05 and then - So my first opportunity was that job, probably in my end of my freshman year, gonna go build this website. And then that opened the whole world of opportunities. It's like, oh wait,

15:16 there's this whole world. I could build this website pretty quickly. I think I developed a sense of like design from that. I've always been that guy. Like I hit poorly designed applications. That

15:27 was my biggest like, like, gotta get that right. It's about the user experience You'll hear that again as we're gonna go get to Pi as well. but I think that started very early. It's like, hey,

15:37 build a really, not only a good looking app or a website, but something that's functional and good looking and finding that kind of middle ground was very important for whatever reason. This is

15:47 like, I don't know why, but it's like probably reading a lot about Steve Jobs and like Apple and like that perfectionism, right? From there. No, but I think it's also, I mean, when you have to

15:60 engage with something that you've built, I think you kind of dog food your way into realizing how important that is to you. How important it is, yeah. So the website, they were super impressed.

16:09 I think I built it in probably a month, month and a half. It was a long time, but for me it was very quick 'cause I had to learn how to do it and then go to actually deploy it, right? And then

16:20 getting a stakeholder buy-in, I had to go show it to a bunch of department chairs and deans and kind of convince them that this is the right approach to the website design, the content, et cetera

16:31 That was a very good experience as a 17 year old kid. kind of going and pitching. What that did was, at the time, our UTSA was transitioning away from WebCT, which was the most used LMS, to

16:45 Blackboard. And so they put me on this team, they're like, go deploy Blackboard for us. And so I spent a lot of time with Blackboard, and I hated it. I was like, this is the worst app from a

16:57 student's perspective. A lot of these apps are really good from an administrator's perspective, because they check off a bunch of features. And that's all that they're doing, right? They're

17:07 saying, hey, it has 100 of these features, maybe we'll use 10 of them, but it has 100. So let's choose this vendor versus the other. And this is true across all LMSs. And that's what kind of

17:18 was the seed for my first attack startup eventually. But it's like, go deploy Blackboard was in the weeds, if you will, of how does it work? How do you go deploy this at scale? UTSA had a lot of

17:32 students, I think it had. 25 something thousand students back then, how do you transition, like kind of being on those meetings, listening to people, how do you deploy something from scratch,

17:42 like that zero to one action, right? So got to learn a lot about that. And I think that job then gave way to, so UTSA had this thing called a quantitative literacy program, where certain courses,

17:56 they were asking instructors to track how a student did at a question level, right? They're like, okay, 30 questions in a quiz, track exactly how they did on each question and compare with other

18:08 students, right? So it's not just like grades from a quiz, grades from a course, question level. Oh, it's a question. And did it track at time it took to answer that kind of thing, or it was

18:18 just. Sometimes it did it if it's on online, it did track how long the student took, but most of of these tests were in person. So it was like on pen and paper, right? But they wanted to see.

18:31 Because what they were trying to check, I think, is a cross-ut. system, how students were performing, if they had a similar test, and then kind of, I think, grade the teaching, right? The

18:40 instruction is how can we improve and give you better resources? So I think that was the idea. But what I saw was instructors were probably wasting 10, 12 hours manually entering this data, right?

18:55 This is an interesting opportunity 10 to hours a week. Yeah, 10 to hours a week. I mean, that could be spent preparing, right? That there itself is time saving that you can spend preparing your

19:06 coursework, right? So you can teach better anyway. So I was like, here's an opportunity that need to make this more optimized, more efficient using technology. So then I started tinkering around

19:18 us like, let me build an MVP. Let me go present it to the department chair and see if he likes it, built that in probably like two and a half weeks, very rudimentary product. He loved it. He's

19:30 like, this is awesome. This is great. This will be super helpful for us. So he's like, let me find the funds to make this happen. But I was like, you know what? Let's make it. Let me make a

19:40 company. I'll give this to you as a service instead, right? Like it's like, I'll give it to you as a service. It'll cost X dollars per user per month. 'Cause that model really works in a tech

19:51 and I learned that from Blackboard, right? That's how Blackboard charges. So it's like, okay, let's go do this. And it worked really well So it started with probably 10, 12 instructors and all

20:02 the way through 75 instructors over 150 courses. It was pretty cool. And then, so it became a real company. I hired a few people. All this while I was still in school, right? So I was 19. You

20:15 had this amazing design partner, right? That was helping kind of you compress this validation period, so you get it really quickly. I mean, it was the same product that you built the day one,

20:27 the one that you used throughout that deployment. I learned a lot by like, once users started using it, we could kind of make it better and better. The hardest part was getting the APIs to work.

20:40 And so we had to kind of get, you know, you know, parse code, right? Parse codes are those OMR sheets. Getting those guys to open up the API. So I can get the information from there and then

20:49 push it into Blackboard, right? So I became the intermediary for this. So that was the app, essentially, right? So completely circumnavigating the instructor having to manually enter that data

20:59 So it's completely automated. Because they already pushed the parse score in, and Blackboard already spoke with them, but it only captured the final grade. It didn't capture the question level.

21:10 So I just built a module that kind of captured the question level and pushed that data out. It was a neat app, right?

21:17 So in terms of user experience, I think it was a very simple app from a user perspective. So I didn't spend too much time there. The user experience will come in the next startup But so it was very

21:29 much very. very simple, easy to use so that instructors don't have to spend too much time thinking about it. That's all

21:37 it was. That's not important, though. The idea was maybe the whole UT system could use this, right? But that didn't end up being. So I think the, I thought the TAM was a lot bigger than it

21:47 ended up being, but luckily UTSA bought it off of me, right? So, and that's what funded my grad school at Columbia University. So that was like my first kind of, and again, very grateful to the

21:58 people back then who supported it. But yeah, all this while I was still still at school. Yeah, school was very easy, Alex. I don't know how to put it. It was very easy. I was a 40 student. I

22:10 was like, this is - it was good. It was pretty easy, but I'm not bragging. It just was - it was easy. So I had a lot more time. I think I was - again, all of that - I owe it to my education in

22:21 India. So kind of school was extremely easy coming in here. I probably already had double - It's funny, I mean, whether school's too easy or people do. I'm thinking of some of the greatest

22:32 entrepreneurs in the world, a lot of them either just didn't focus on school, whether it was too easy or too hard and didn't care or whatever it was. It almost feels like if you zag away from

22:47 school, it's created some of these amazing entrepreneurs because they were able to focus on the things that they were actually passionate about, develop a lot of skills around that That's why I also

22:55 love America. I think especially at UTSA, if you are willing to work hard and they saw that I was. I did extremely well in school. Any tasks they put in for me, I did really well. They gave me

23:09 more opportunities. So, yeah, extremely grateful to UTSA in that sense. The kind of projects that I got to work on, so in my last year in college, there was this open compute institute that was

23:23 set up. It was a collaboration between Facebook, the ARM processor, UT, the UT system. So it was called Open Compute. And I was the only undergrad in that group. It was all PhDs and I was the

23:35 only undergrad. And I got to kind of set up these servers, right? And just work hands on it. That's awesome. Did stuff that I don't think people have, I'm very grateful to that, right? I think

23:46 that you learn a lot from it. Well, you also built your way into it. Yeah. Yes, you were given that opportunity, but I think you, what is it like? Like luck is where preparedness meets

23:55 opportunity. 100. I mean, there's a little bit of that. Yeah, yeah Maybe a lot better than that. Yeah, a lot better, yeah. But yeah, I consider myself extremely grateful. And, but I

24:04 learned a lot through all of that and just interacting the best of best. I was reporting into the VP of Rackspace back then, right? So he was, he was like a mentor for me. Like just kind of gave

24:15 me a lot of pointers. But he also helped me kind of find to my business mindedness, right? He's like, you're very technical, which is great. But if you really want to go far, you got to think

24:25 about it from a, Like, ultimately, where does this fit in a. It may be the best thing that you're building, but who's willing to pay for it, right? So that lesson, I think, he gave me. It's

24:37 like, I think, kind of continuously think about that. So did you go straight into grad school from undergrad because of this? Okay, yeah. I did, so got into Columbia and a couple of other

24:48 high-be leagues. I chose Columbia because it was in New York. I got into Rice, by the way. Rice, I got into the PhD program at Rice I was very tempted 'cause also they laid out the red carpet.

25:00 They flew me and they kind of got to come here, et cetera, but ended up to go into Columbia for New York City more than anything else. Also, I was from India, so

25:11 weza, right? That was playing on my mind. In fact, I was the first student at UTSA to use my student weza to start a company. They didn't even know how to do it. So I was like, yeah, it is

25:20 possible. It's a

25:23 law, like you can do it. Right, so - Is a student visa limit used? I remember when I was working in Australia on a student visa, it was like, I couldn't work more than 20 hours in theory. Yes,

25:32 so that is true for what is called pre-OPT, right? But you can use your OPT, which is optional practical training for, to be self-employed, right? And that's a one year, one year visa, and if

25:46 you're on STEM, that's up to three years. Oh, that's right, okay, okay. So I use that, but then the paperwork for, if you actually got an employee by a regular employee is different from then

25:56 self-employment. So nobody at UTS had done that. So they're like, is it even possible again? It's possible. Like, so - I spoke of all the other student visas at UTSA that have gone on to start

26:10 companies. Yeah, it's from there, very first. I'm sure now the advisor knows, it's like, hey, if you're interested in this, this is a path. It is possible, it's illegal, right? That's off.

26:20 And that's why Columbia happened I was like, okay, candidly, I don't think it's when. too many classes at Columbia. Because I'd then started my second company, which was Edulink. I was a

26:30 collaboration tool for students and instructors. The idea with Edulink is where the user experience comes in. I was very, I

26:39 think I hated the user experience with Blackboard and then Canvas, which was the other LMS. I was like, how do we build a product that puts the student at the center of design, right? Kind of

26:50 make it easy for students to be able to use this tool, meet them where they're at, right? And that becomes a theme with the rest of my life, but it's like, hey, meet the user where they're at,

27:03 build it around them, like what is it that they care about, right? So what do students care about? Good grades, right? An easy, easy way to get those grades, right? Not only just good grades,

27:13 but it's like, hey, minimum effective goes towards a good grade. Exactly. So the way that Edulink was organized was

27:21 There's an assignment on any of their LMS. We pull that assignment and then we found content that'll help you solve that assignment, right? So it's like YouTube videos, other articles that are out

27:33 there. So it's like a simple recommendation system that pulls content from the web by crawling the web and helping you solve that assignment. What are that assignment? So we had to kind of

27:42 understand what the assignment is. The first perplexity. Yeah,

27:45 yeah, this is somewhat very rudimentary So I wouldn't say because I didn't get to kind of that scale, if you will, and I would have loved to kind of get it to scale. But it was, hey, here's your

27:56 assignment. Here are all the resources to help you solve it. One, two, based on where you were sitting on campus, it found you peers who could help you with the assignment. Oh, cool. So it's

28:07 like Uber for tutors. Yeah, yeah. It's like, hey, I'm around, I'm in the library. I can come and help you. How did you get the profiles of - These were already tutors that - So students can go

28:15 and kind of create profiles and say, hey, I'm an expert of x, y, and z. or I've taken these courses in the past. So hence I have knowledge of how to solve them. So that was the first iteration

28:27 and then we had like a slack like collaboration where you can kind of create channels with your peers and discuss different things. All, I think the hardest part is was consumer product, right? So

28:39 I was like, I don't know if I have the selling chops to sell a consumer product. A, I was not part of any frats. So I didn't have that kind of like a market if you were. Right, right The

28:49 distribution, so the only way I could think of kind of selling this was go and sell it to the university, top down, make it easily available, right? And make it available as a tool and then hire

28:59 ambassadors to go and kind of distribute, right? So that was the idea. So to apply that UTSA, Columbia University, a few classes at UT Austin as well, use them. And then I

29:12 was trying to raise, I think I'd move for the most part, by myself and then friends and family, like early angels, like a lot of angels in the, in the tech space had funded it. Cause it's seen

29:21 that I'd done well with the previous one. I think, yeah. So New York City, again, very kind to have that kind of a tech network. So I'd known the who's who of, of a tech in, in New York. And

29:33 I was trying to raise a larger round, whatever you want to call it, like seeds, seeds, say somewhere in between at the time. And the CEO of, of, of Pragya, which was a publishing company out

29:46 of, out of Boston, he, he wanted to put a check in to the company. But then he spent time with me and, and the rest of the team is like, looks like there's a lot of synergy in terms of what

29:57 we're doing. We just want to acquire you, right? So, and I was like, I loved him. And his co-founder, the CTO, was the VP of WebCT, VP of Blackboard and VP of Experience and Engage, which

30:13 was all these best attack tools back. then. So I was like, this is a great opportunity. Let me, so me and my team then moved over from New York. What, what about the first acquisition process

30:23 and form the second one? What change? I think the first one was, um, so the code had to have MIT license packages, right? So I think when I used a lot of those packages, they were not open

30:35 source. So I literally had, I think, a week and a half to go and rewrite some of the code because it had to be a MIT license. Otherwise, they were like, Hmm, we don't know if we can take this,

30:45 right? So it's a second time around, very intentional about which packages are we using when we're building the product. And then this, I think the second one, like I said, the emphasis was user

30:56 experience versus, Hey, let's build everything under the sun, right? It's like, let's build three things and let it be the best things that we build. That's probably learning from from

31:07 Blackboard and my first company. Second company, I'd say, I also focused I had some really interesting people join me early on. Unfortunately, it didn't work out. I think a lot of them were

31:21 immigrants. They were worried about Visa. They were like, you know, they got offers from like Palantir and like Microsoft. And so they ended up kind of choosing that path. So then how do you

31:32 kind of get them off the cap table? Like a lot of learnings associated with that.

31:38 You're attracting some incredible talent Yeah, I think I can talk to the vision, right? I think and then once people rally behind the vision, and these are really smart people, right? So they're

31:50 like, yeah, you know, let's do this together. It makes sense what you're saying. I'm sure they feel the pinch as well. And they're looking for other people who want to go solve it. So I think

31:59 edge link was just that. It's just finding those kind of people. I love working with people who are very much biased for action and have high ownership, right? And those are the people that you

32:08 want to surround yourself with, right? like we don't want people who want to be spoon fed all the time. Um, and so from, but from a technical perspective with edualing, I think one of the things

32:19 that I learned is how to competentize your code base, right? So basically build it in, in, in blocks that can be reused and repurposed, right? For different things. So very much the thinking of

32:32 how do we be very efficient with our architecture, right? So that way we build it in blocks, uh, and then reuse these blocks for any other new ideas, because I had so many ideas, but then don't

32:42 want to boil the ocean on day zero, but still build keeping that end golden in mind, if you will. Right. So I think Edulink taught me a lot about that is how do we build these components that can

32:53 be reused, could be UI, could be back end components and set ourselves for efficiency, right? Um, I think we, we hired about four engineers, one designer, uh, and one QA engineer is a very

33:05 small, small team, uh, but still all of us were extremely hungry We wrote a lot of code, the first. So the first, with Agilink, when I deployed it at UTSA, with one of my computer science

33:19 professors, it was 55 students logging in. And I think we'd never tested it with more than 20 people at a time. I was like, and then the verification email goes out, right? Like to verify your

33:31 email. I was like, oh my God, we never tested this with more than 20 people. It's like 50 people suddenly now. Let's see how it goes. But it went extremely well. And I was sitting from the back

33:40 of this computer lab and watching all of them do it And their eyes lighting up as they saw the product. And that feeling, Alex, is what I live for. Is when people's eyes light up and they're like,

33:53 what is this? This is interesting. I want to use this out there. And then they tinker with it by themselves without you having to tell them what to do. Then you've won, right? I think then your

34:03 user experience has won. And then it has to move fast. If they're stuck on a loading state,

34:12 So even if you're doing something that takes a long time, just make sure you take them to the next step, keep them engaged. So that was, again, all the engineering. That's where the engineering

34:20 went in. It's like, we had to pull these recommendation videos that took some time, show them something else in the meantime, to kind of keep them engaged. So learning a lot of how do you kind of

34:31 architect your engineering product with low latency? I think that's, again, a

34:38 lot of lot of learnings All of that was self-learned for the most part, right? But yeah, that was - I feel like it was a digital manifestation. I'm sorry, I'm - No, no, this is fascinating to

34:48 me. I don't, you know, I feel like this is a digital manifestation of every experience you've had before that. Yeah, 100, 100, I think it is. And you see that I think a pattern developing,

34:58 right? So like there's like a user experience, efficiency, building things modularly and kind of reusable components, very much with, hey, build an MVP.

35:12 get it out there, get some feedback, and then build on top of it. I think all of those things I learned and applied with my second company. And that's what I think got the Prager Guy really

35:25 excited. It's like, this is the culture that I wanna develop, right? And maybe this kid can help out. I'm still a kid at that point.

35:34 And so you've gone out through two entrepreneurial cycles. Yeah. And then obviously you were working at Prager afterwards, but then switched to toast. Yeah, so actually between Prager and toast,

35:45 I did a couple of startups. Oh, okay. I was like, like the last few years, because I had completed, I invested my shares, I'd built like seven products for them in the span of six years, right?

35:55 All zero to one action, talking to universities, deployed it,

35:60 but I started two companies. One was a blockchain company in India. So this was, hey, land registration is hard. Can we put it on blockchain? Built all the tech. We need that here in the US.

36:11 by the way. Yeah, it's hard, man. Like I think getting the government to buy in was incredibly hard. I had a really good investor in India. He was very well connected. He put in the cash, he

36:22 put in about 250 K. He's like, Hey, here's, take this money. I trust you. I really like you. I think you're a smart dude. Go figure it out, right? So, and then he connected me. So all this

36:34 while I was at Pragueia and tried to do it, it just was extremely hard to get buy-in from the government. Unfortunately, that the product peated out, but learned a lot about blockchain, right?

36:47 Like distributed systems. I think that was my first foray into distributed systems and how do you architect distributed systems, right? So, a lot of learning through that. And then I started

36:57 another company in 2020. This is just as the pandemic hit. It's called Envy. It was basically a creator,

37:07 platform like for, for courses. So have you heard of Maven? And a lot of different contexts is what, what's this? It's like Udemy's, the Udemy's go

37:16 from the start of Maven, where he's just trying to get these like experts from the industry to come teach courses. Yeah. Right. So it was a very similar thing, masterclass type, like a

37:24 masterclass thing, but for anybody can sign up and, and go do it, right? A lot of learnings from that is how do you create like a video platform where people can stream live and, and make that

37:37 whole user experience very simple to create those courses, because these guys are, they don't have time, right? They're, they have like a full time job, but then you have to incentivize them to

37:47 go and build this course and make it easy. So very much the emphasis on user experience. And the good news there was Alex, I built a lot of components all along the way, right? So these

37:58 components were being reused. So I built a website builder, a simple landing page. I'd build a course editor with edulink. use that in edV. So a lot of those, and that's why - And you always

38:09 modularly. So you always had this - Always had these microservices. And the, of course, the front end was very different. The UI was different. So we had to build a lot of that from scratch.

38:18 But the back end, the concepts were there. It's about kind of dusting off and kind of repurposing it. But that's why I'm fast, and that's why people think I'm fast. It's that, right? It's like

38:29 that modular thinking and building things with keeping the future in mind and knowing that this may be useful later. Let's build it modularly. Versus, I think oftentimes people build it like, okay,

38:42 let me get it out, and then I'll figure out the modular part. I think I was very clear, very early on, so let's build it modular and reuse it. Yeah, the game at the end in mind, kind of. Yeah,

38:50 yeah. Not always the best strategy for a startup,

38:55 but I think it worked out for what I was trying to do, right? Because also these modules made sense. Like one was like a payment gateway, right? Yeah, like you can feed it a course,

39:06 a land transaction, you can feed it POS transactions. Pi comes, comes to mind, right? So, but it's, it's, it's a payment gain development. And it talks to Stripe, it could talk to Razorpay,

39:16 which was in India. So it was very modular in that sense. So that's the idea, right? And the same with the website builder. Website builder is a website builder. It can be used in the course

39:24 context. It could be used by a university. It could be used by a regular creator. It doesn't matter, right? It's a website builder at the end of the day. So that was the idea. I think in 2020,

39:35 that was more like a passion project than anything else. I saw that education was gonna move away. My, again, I still believe that education is gonna get extremely personal. It's like, because

39:46 again, with AI, it's like, you learn differently and I learn differently. So it's gonna be very personal. So, but along the way, interestingly, I'd built a bunch of fintech products, if you

39:56 will, right? It was like, A, with Progia, I was creating demand for universities SMB for dimension is Pi? right, Dementia..

40:05 Um, and with, uh, with the Edvie, I was building a payment gateway to kind of process payments, uh, not at the scale that of course, toast, et cetera did, but I still got to build something

40:15 from scratch, uh, and then learn through that, and then I can talk to it, right? It's like, okay, these are the problems that we face. So I think people appreciate when you talk to the

40:21 problems that you faced. Um, and then I got tapped by toast a couple of years after. So they had a new one, just program before, before we get to toast, I have a question around the education

40:32 piece. No, because I, I mean, I've got young kids And, and I, I, I kind of see a little bit of, of what they're going through. They're private school and we're blessed enough to, to do that.

40:42 Um, but they, uh, I feel like their zone of proximal development is, is different than the, like the minimum, the common nominator, like approach. And it's just like, that's just how we've

40:52 taught for such a long period of time. We've got one, one teacher, one professor that's, you know, basically trying to, you keep up with 20 students or more, um, and, and leveraging a tool

41:03 like AI to, to get in approximate development to where it meets the student to where they need to be pushed a little bit more. I agree is the future. How do you see, given your experience there,

41:14 and then of course, what you've been able to build at PI, how do you see education? What do you think education looks like in five years? I don't know about five years. I think five years could

41:22 be very similar. With education, it's always a lager. It'll take time. I think more tools will be available for personalized learning. So I'd say most of the learning will happen outside of a

41:34 classroom, right? That's my thesis. So when you go to the classroom, it'll be more about, let's discuss what you learned from the chat GPT conversation that you had, right? So make it more

41:44 about a discussion, versus, okay, let's sit and read this together, right? Yeah, it's like your hour at the computer and then let's go take it back. Let's go take it back. And so you be

41:54 personal, right? And so that way, what that does is, not all of us learn at the same pace, right? Some of us take more time. Some of us learn things very quickly. So you can kind of do it at

42:04 your own pace, that learning bit, right? But then all of us come together in a common place and then discuss what we learned, right? So smaller groups of discussing and then hopefully learning

42:16 from that and kind of either reinforcing what you've learned or maybe changing your mind about, okay, maybe this is not the right way to think about it and here's another way to think about it,

42:23 right? So I see that being the future. The second thing is, I do think our courses today are very much like, it's a computer science degree,

42:36 AI degree, whatever, ML engineering, et cetera. I think it's gonna be more skill-based. It's like you go pick up these skills, almost like the dual lingo. So go pick up these skills and then go

42:44 make of it, what do you think, right? Okay, good build. So it's gonna be more skill-based versus credit-based. That is the future, is that eventually you can pull together a bunch of skills and

42:54 that becomes quote unquote a degree, right? I love that, yeah. But you're like, it's like this Google map, You're going to go figuring it out. You pick up skills on the way, right? And then

43:03 it doesn't have to be computer science. It could be, I don't know, like it could be ML plus economics plus finance accounting. I don't know, like it's stuff that you're good at and that you enjoy

43:13 doing. I think that's going to be the majority that you still are going to have your traditional degrees because we still want research, right? We want to structure research. Otherwise, you're

43:22 not going to be able to create the PhDs who actually do some of the zero to one action there. So we still need that. I'm not saying it's going to get replaced. For a majority of us who don't really,

43:32 I think we take these 120 credits at college just because we have to versus we want to. I think we'll have the luxury to choose what we want to study and make our own degree, if you will. I see

43:45 that as the future in IT. It makes it so. It reminds me, you mentioned Steve Jobs. It was at the commencement season at Stanford, where he's like, you can't connect dots looking forward. You

43:53 can only connect them, you know, looking back where it's in hindsight, right? And so being able to take those data points, and then I don't remember who to attribute this quote to, There's a

44:02 couple of different ways to truly find your passion, right? Because there's a theory that like passion doesn't really exist. It's just like what we think we're good at. And so it feels good,

44:11 right? And so one is to truly be the best in the world at this thing. Or to be, you know, take two or three things and uniquely put them together in a way that nobody else has. And so what you're

44:22 describing is a little bit of those elements, you know, conflating into something that's unique to you as a person. That's amazing. Exactly. I think that it'll feel you, right? Like like you,

44:30 and I think we'll create a better world, happier world. That's my hope. Yeah. That's my rose color glasses. There you go. But that's, that's the hope, right? And you never took them off.

44:40 Yeah, I don't think we should, right? I think we can live life with those on fully knowing that life doesn't always pan out of the way we want it to, right? I have this quote, you'll probably

44:51 hear this multiple times from me. The universe gives you what you need and not what you want. Right. So it's, it's, it's just that that's life, right? So, but you should still work as if you

45:00 will get what you want. you'll get what you need.

45:05 That's the best way to think about life. I'm hopeful about the skill-based education system. That's what I would love my children to grow up in. That's where AI is really useful. I

45:18 would say hard pivot to toast, but no. Your whole life has been built with these same threads, these same core values of a user experience, of efficiency, of modularity. While toast is a

45:34 different industry, I imagine a lot of the same aspects. Same aspects. I think the reason why. I had job offers from three companies at the time. Toast had this new wenches program. It was a

45:45 startup within a larger toast organization.

45:50 My boss at toast, I think I got lunch with her. I really liked her. I was like, she was an ex-entrepreneur. She's done a couple of startups. Pretty big exits. I think we just got along, and I

46:03 think the way we looked at the world was very similar. So the other two offers, in fact, like, monetarily were a lot better, right? I was like, I need to be here. For whatever reason, my

46:14 intuition was, you need to be here. This is the place to be, and I haven't met Syed by then, right? Like, none of that has happened, but I was like, I need to be here. This is the culture

46:25 that I want to be a part of, her culture, not so I hadn't met. I mean, I knew of off-toast, hadn't met anybody from toast So I didn't know much about it, but I really enjoyed talking to her. I

46:35 ended up joining the New Wences program. So the first task that they gave me was, here's this product, go build it. I'm like, OK, I'll do it. So I became the designer first. So I designed the

46:47 product. Like, oh, you design them like, yeah. I'm like, whatever it takes. So then I was like, can I talk to a few customers? They're like, oh, you talk to customers as well? I'm, yeah,

46:54 I can talk to customers. So when they spoke to customers, they - their eyes lit up, when they saw those designs. They're like, yeah, this is exactly what we want. And that became the first

47:02 product that we launched Benchmarking. Okay, yeah. So Benchmarking was helping restaurants understand, kind of look outside their own four walls. So let's say they're selling croissants, right?

47:14 How are, how are microsons priced versus croissants outside, right? Is there room for me to maybe charge more, charge less? Also like compare myself with my peers. Am I doing well? Is it like

47:27 seasonal? Is everybody doing poorly or is it just me? And then taking certain actions from it. So it was the first AI product for toast. So we had to train, like that was such an interesting

47:38 project because menu items across the board are very different, right? So what a latte at one place could be something else somewhere. It could be cafe latte, right? And so we had to kind of use

47:48 AI to say, okay, both of these are lattes, right? Or somebody would say burger, right?

47:55 Like, why no one like, but no, but again, it's, it's not like the nomenclature can mean different things. Exactly. I mean, it's, it's a creative field. So we had to kind of get everything

48:05 standardized. And so that was like a machine learning exercise. So we'd come up with some training data where we manually went through like

48:13 20, 000 rows of like many data and then classified it. So we, it was a team activity. Everyone did it. Everyone got like 5, 000 rows that they had to go and kind of classify And then we had this

48:23 data science team build a model for us. So that was a very interesting experience. And then we launched that in record time, I think it was like probably six months. Yeah. So six months. And it

48:36 was, it was just all, and then people love it. People love it. The attach is so high. It's, it's crazy. So we had a free and a paid version. It also touches into like a basic human behavior of

48:47 like, I want to compare myself to others. I don't want to know how So it worked really well. I think gave me a lot of street cred, ultimately say, Okay, this guy can build stuff, right? The

48:60 second product was an ads product and ad server. So we were using toast inventory to kind of showcase people like AMX Coca-Cola on toast inventory and paying some of it to the restaurant. Just that

49:13 scale. So I think we were driving probably like a billion impressions a month. So building that for that scale was a lot of fun

49:22 Yeah, so I think I loved toast. When I joined, we were three engineers. By the time I left, we were 15 engineers. So I got to recruit like a kick-ass team. Yeah, yeah, yeah. And I enjoyed

49:33 that process of just like meeting with people, figuring out if they're a good fit or not. I had never recruited such a lot. What about them, especially having recruited as such a great team,

49:43 Nejilink before, what about them and other startups? What about them, did you say, this differentiates them as an A player versus a B player? So whenever I talk to people, I look for a couple of

49:54 signals, right? So first signal is, hey, tell me about a time when you, when there was a problem, right? And what did you do about it? Oftentimes people are 5050. They either are able to talk

50:05 about, here's the solution, but somebody else identified the problem. Here's the problem and they'll rant about the problem, but not do anything about it, right? So for me, that A player is

50:15 they identify the problem and then actually care enough to do something about it, right? It's that shows bias for action and then the second thing is given very little information and oftentimes my

50:26 questions are very open-ended purposely. It's like go figure out and ask me the right questions, right, to then get to the meat of the problem. I'd say only 10 of the people that I talk to are

50:40 able to do that well, right? So it's like get to the middle, especially in Genome and I'm talking about engineers I did get to hire some awesome designers and product managers at TOS. That's why I

50:51 love that role-man at Toast, because I was a leading engineering team, but I could do all of these things, and my boss there was, she was just awesome, she's like, yeah, do whatever it takes.

51:02 But anyway, going back to your question, it's about, do they have high ownership, and do they have high bias for action? I think these two are core qualities for me to have fun working with them,

51:13 and for them to have work with me. Otherwise, I think that relationship will probably not work out in the long run. Not that people who don't have it are not successful. I think it's ultimately,

51:24 when you're building a team, you want like, you're going to be spending eight hours a day with them, right? You want to have fun along the way. So I think it's about choosing these two are I

51:33 index very heavily on these two attributes, and if they don't have it, or I don't get positive signals for any of these attributes, then because I don't care about, hey, I can write the best,

51:44 like, program, right? Or like, you know my code is very clean it's it's it's awesome it doesn't matter to me like it Today, with AI, I can write better code, right? So that doesn't matter to

51:55 me. To me is, are you able to take and run with very little input, right? Or is that you require a lot of input? And then are you taking feedback really well, right? As you get this feedback

52:09 along the way. I index more heavily on those attributes versus, oh, like here's a coding problem, 60 minutes to go solve it. Let's, like, I wanna see if you solve it in the most perfect,

52:20 efficient way. 'Cause I think all of those other things can be trained, whereas these qualities, like bias for action, high ownership, feedback, those things are, they take long, they take

52:31 long to train. And it's better to pick people who already have that versus versus not, right? And the team that we have, at least you wouldn't believe it, I think all of them, they bring me at

52:42 like Saturday night at 2 am, I didn't ask them to work on it, right? They're like, I'm working on this, Like, how does this look like? Nobody asked them, right? But they're doing it. And

52:55 this is, they're working at Toast, right? It's not like they're working for a startup too. And off like leadership at Toast was asking me, they're like, how are you getting them so productive,

53:04 right? Because we've seen, and a lot of these guys were also in India, we've seen like teams in India not performing as well. How are these guys so productive? I'm like, I get out of the way. I

53:14 give them direction and get out of the way. I don't micromanage there. Well, a lot of the effort to understand them has happened at the forefront. At the forefront. And then you don't have to

53:22 worry. Yeah. And like kind of set the vision. This is like so you too, right? Like you build and design in the architecture. Everything is so detailed, oriented around like the user experience

53:34 and the efficiency and modularity. All that takes time on the front end, but everything else becomes easier afterwards. Yeah, I mean, it's like probably me playing cricket helped because the

53:44 preparation is before the game, right? Once you're in the game, you're in the game just like go with the motion. There's not much, I typically don't get upset in the moment 'cause there's no

53:55 point, right? It's always either a post-mortem or prepare, right? Prepare better. Maybe I just didn't get the right instructions going in. So even if things have go south, which they've gone

54:06 multiple times, there's no point in getting upset about it in that moment. It's like, let it go, it's okay. We all make mistakes and then come back to it. I think I'd say the team appreciated

54:17 that a lot. They probably didn't see that with other leaders, especially in India I'm talking about India. Here too, I think I had some really senior engineers reporting into me, much older than

54:27 me, right? But even at Pragya, right? One guy was like in his late 40s reporting into me, right? So, but they were, I think for me, it's like getting, exciting the vision and getting out of

54:40 their way, right? That's the best way to do it. Not micromanaging anybody. And always intervening if they need help and figuring out where they need help, doing a pulse check Have this one on one.

54:51 planning sheet that I use where it's basically four columns. First column is goals. Here are my goals and here's how I'm going to achieve them. The second column is moonshots, right? These are

55:02 moonshots that if I take, I'll get better and it'll also benefit the team, right? And so we spend every month, kind of just go reviewing this. And then the third column is learning opportunities.

55:12 And then the last is learning resources. Like what are the resources that I have to be able to attack the first three, right? So that spreadsheet, so rather that Google document is something that

55:23 I use with all my one-on-ones. This is something that I've been doing since the edge of link days, right? I think it helps put things on paper, right? And then you use it the way you want to use

55:34 it. Like if it's not working for you, that's fine. We don't have to use it. But 80 of the people tend to use it. They like it. They updated themselves. It's not like I'm pushing them. It's

55:42 theirs. And then we use that to have our conversations. And that's how you grow them, right? ultimately, you grow them without explicitly growing them. So, yeah, I love that. Okay, so how,

55:54 I mean now at PI, right? Yeah. And building PI, some of those ideas started within toast, I imagine. They did, so I think the other aspect of my role at toast was I got to talk to merchants

56:06 almost every day, right? So I was talking to a new restaurant every day, just asking them what their problems are. One thing became very clear is they don't care about technology, right? They're

56:18 like, I don't care Call it AI, call it BI, CI. I don't care. It doesn't, to me, it doesn't matter. I'm a chef, I wanna cook, right? And I think this attribute is true for any SMB. It's

56:30 like they're so passionate about the work that they do. Everything else is just like work for them, right? It's in the way. Exactly. And so the more work you can take off of their plate, the

56:40 happier they'll be 'cause they get to do what they love more, right? I think that core value came out of my conversations, talking to a bunch of restaurants is like, and then the other The other

56:51 thing is a hair and fire problem for them is, okay, I have all of these SaaS tools that you're helping me, like benchmarking is great. We had like a different marketing tool, like helping

57:01 reactivate existing customers, but help me find new customers, right? And that was a hair on fire problem. And they were

57:09 paying agencies thousands of dollars a month

57:13 to do that. And they're like, I don't understand any of this. I don't understand what clicks and impressions and all of these things mean, like my bottom landing growth. Yeah, I don't care.

57:23 Like I put in100, how much did that100

57:25 make me, right? That's all that they cared about. I think that came out of my experience talking to, talking to, also met Syed during the same space, but I think he just came out of - Co-founder.

57:37 Co-founder at Epi, probably the best sales guy that I know and he's awesome. He used to call him the magic wand on our team. But I think a core value of, Hey, I don't have time for. thinking

57:50 about all of these things, can you take it off of my plate? Like this marketing, finding new customers. But it's an important problem. I don't know how to address that problem. And that became

57:59 ultimately the mission of Pi is how do we help customers grow and reduce costs? So if you look at our Pi's vision statement, it's exactly that. It is helping small businesses grow and reduce costs.

58:15 And those are genesis of Pi, I would say I think I also got to spend quite some time after with the Google team got some friends at Google and kind of, when I first presented the architecture for Pi

58:28 to them, they were like, this is an interesting way of using our APIs, right? Yeah. But that's then back from like you having to figure out how to tap into, you know, the original edtech type

58:33 APIs, right? Way back in - Way back in the - Blackboard and yeah, yeah, so it's it

58:48 And not only, I think even at Pragya, because we were doing demand-gen there as well, is how do we create demand across the board? How do we tinker with things and get to the root of what they're

58:59 doing and then use them slightly differently from what they were meant to be used for, right? Sometimes you get dinged for it, right? Luckily, in this case, we don't get dinged for it because

59:09 from Google or any other of their perspective, it's like ad spend is ad spend. The money is coming in, it's just so you're optimizing certain things and making it useful for yourself. So they're

59:20 happy that we're doing this because we're gonna get them a whole new set of customers that previously never advertised. Yeah, I mean, Google's never been great about quantifying what the ROI is on

59:29 that spend, right? It's just like, oh, they became table stakes so quickly. It was like, I know we have to do it. We have to do it, yeah. Because everybody else is doing my competitors you're

59:36 doing it, but I don't know what that's manifesting in terms of ticket size, account of customers that come through my door. Exactly, and this is not only true for Google, I don't think we should

59:46 beat Google down, but it's like true for all of them. I'd say like any of these other ad partners, Google, Meta, all of them that are out there, the user experience is not the first thing

59:56 because that's not their bread and butter, right? For them, it's about the algorithm that finds you the right customer. That's what they're indexing on, which is the right thing to index on for

1:00:04 them. But from an SMB perspective, a small business owner who is not able to pay thousands of dollars to an agency to set it up for them, right? For them, it's like, no, I need an engineering

1:00:15 degree now to figure this piece out, right? So can you help me with that? I think that became the genesis of the idea. So we took a process that typically takes about four to five hours and

1:00:27 distilled it down into five clicks, right? So they clicked five things and they're up and running. And this kind of goes back to the, what I think is an AI-native company, right? I think an

1:00:38 AI-native is not writing code with AI.

1:00:43 I think AI Native as a company, and this advice is only for startups, right? I want to caveat that this may not apply to larger companies. I think for me, AI Native is, it's about building

1:00:54 intelligence and using intelligence, right, across the board. So intelligence could be a product spec that I came up with after talking to chat GPD, but then a product spec by itself is not useful,

1:01:04 right? How do I take that important to cursor, right? Like how, how can I do that piece? And that's what we're working on right now How do I make that better and better? How do you say it, take

1:01:13 the same product spec, give it to a designer that she or she can use in Figma to create the design, right? Once the design is ready, how do it, then I feed it back to cursor. So it's ultimately

1:01:22 intelligence that's moving, right? And then we're using AI to just augment it, if you will. Got it. It's a reason through that entire process. Exactly. And so as opposed to having an individual

1:01:32 take the data, move it to one tech, one platform, then to the next one, as it's being adjusted, it's happening organically through reasoning. more organically. I think we still haven't cracked

1:01:43 that piece yet. So one of the problems that I often use, I talk to Chad GPT quite a lot. In fact, I use Chad GPT to prepare for this podcast because I was like, here are all the thoughts that I

1:01:53 have. Now let's talk to each other to figure out what I want to say and how I want to say it, right? So very much an AI native CTO for you. But the idea is that, okay, there's AI, Chad GPT

1:02:06 spits out, like, let's say a product spec. But how do I make that actionable by cursor, which is what we're using to build, like, right code, right? So it needs to be in a way that cursor can

1:02:15 understand it with the context of off of the code base. I think that piece was still perfecting. We're not, we're not there yet. And then briefs to designers is like, how do I give a brief to

1:02:26 designer? So he or she can go design, what is it that we want to want to design? So but ultimately the, the net net point there is its intelligence, It's that's what I need a ways is like, how

1:02:37 do we build intelligence? and then we move that intelligence around, right? So use AI for everything except for a vision. That's the best way to think about an AI company, AI native company,

1:02:47 right? So vision is you, everything else is AI, right? That's where I would love to get us to. We're on that path, we're doing extremely well. And in order to do that, Alex, and this is

1:02:59 probably gonna be controversial, I think as a startup, the first thing to hire, especially if you already have a technical person who can quote and write, that was me

1:03:08 I think your first hire should be a designer. Oftentimes, it's startups. Designers are one of the later hires, right? Like you have your three engineers and then you hire a designer. On our team,

1:03:19 we have three designers. One engineer, I mean, I hired her a couple of months ago, but it was three designers. And so what happens is you're making mistakes on Figma versus in code, right?

1:03:31 Because today with AI,

1:03:35 effectively code is free, right? It takes no time. So what is scarce is judgment, ultimately. It's like making the right thing. Anyone can build anything. It's deciding what not to build,

1:03:50 right? And so that decision can happen once you put it on paper on a Figma file, like this is not what we want to build, right? So we're making more mistakes, quote unquote, during our design

1:04:02 phase, versus making it during the engineering space. So I have three designers running on three different problems They're building all of these things. And then the engineer, that is me and our

1:04:13 other engineer, we just basically cycle through, right? We have these like sprints. So one sprint, we're working with one designer, while the other designers are working on new design, right?

1:04:21 And just cycle through and go building. And that's how we're so fast. Because the designers are - and I'm working with these designers every day to figure out what is it that we want to build. How

1:04:30 do you want to make this user experience simple? And that's very important for us, right? We don't want to put a product out there that's not simple. We got to go back to the drawing board. Is

1:04:39 this analogous to like a four deployed engine? And like almost like an internal four deployed engineer that's someone that is like understanding what the customer could want or does want and building

1:04:49 an architecture that then gets built on the back end? I see, I'm playing that role right now. I

1:04:56 think what I'm trying to eventually envision is a merging of two roles, a product manager and a designer. They become one. So we hired very product-minded designers who are able to talk to

1:05:08 customers. Today they sit with me and we do it, but in the future I see them just running by themselves. They're product managers effectively, right? And they're designing and then what they

1:05:19 produce is what the engineer consumes to then go build it, right? So our first three hires were actually designers and then an engineer in addition to me, right? And I think that is the model to

1:05:31 follow with startups today who are AI-native because anyone can build anything. So it's like very important then is what are you building and how could the user experience is? And if there is AI,

1:05:46 if you're using AI in your product, I think that chat is the lousiest way to interact with AI, right? Is helping using AI where the users are at. So the reason why cursor was so successful in my

1:06:01 opinion is that it met where the engineers were at, which is the IDE, the code editor, right? Like, imagine if I had to go to chat GPT, talk to it, it produces code, copy that code, put it

1:06:12 into my ID. I'll use it maybe 10 times, but then after that, I'm like, this is too much work, right? But the reason why cursor and other tools like that were successful is because they met me

1:06:23 where I'm already at. I'm already at my editor, right? Very comfortable being there. Now I'm documenting it with this AI conversation that I'm having. And with a click of a button, I'm able to

1:06:32 apply changes, right? So that's our philosophy with design for SMBs, we're very much thinking about meeting our users where they're at instead of expecting them to use chat to interact with us,

1:06:47 right? Because the thing with chat is that it expects, or AI chat is that it expects people to know to ask the right questions. It's hard to ask the right questions. Okay. Right. You're doing a

1:06:57 really good job with it, but I think it's really hard to ask the right questions. And we want to take that out of the picture. So we're, but we're still using AI, right? We're using it to

1:07:07 generate copy. Uh, we're using it to make them more efficient. Uh, we're using AI to kind of just make the whole process efficient. And this is like constant optimization of campaigns. All of

1:07:17 that is AI, but from a user's perspective, nothing has changed. Right. Does, does the role of the engineer change in this new structure? I, I think it's, I guess it is a new structure, but

1:07:29 it's definitely a new hiring plan. I think we're definitely going to be hiring more product minded engineers versus, um, at least early on, right? Now, as this scales, the biggest bottleneck is

1:07:39 gonna be infrastructure, right? And by the way, no big company has figured that out even today, right? It is like in a world where everything is AI-native, right? We have so much data. How do

1:07:50 we not hallucinate and like gonna continue delivering an awesome music experience with so much more data? We people at like big companies are still struggling with that problem. We don't know what

1:08:03 the world's gonna look like in the next five, seven years It's a very interesting space. It's an evolving space. But at this scale, where we're at right now for hundreds of users, probably

1:08:14 through 10, 000, 20, 000 merchants that we'll work with, this I think system works, where we have designers. And then what I'm envisioning is each designer is gonna have one tech lead and one

1:08:25 engineer. So it's like a two engineer pair with that designer. So they're a part that they run on one problem, right? Ultimately, it's a problem that they're solving and using technology, AI to

1:08:34 solve that problem So that's how we're.

1:08:37 Positioning ourselves, we'll have three pods because we have three different products and then each of them are running on one problem thread. I love that. Yeah. I love that. This is so neat. I

1:08:47 know we're like 1030, so I just want to make sure we're good on time, but I. I'm sorry, we've been. No, this was too much time on my journey. I was super fascinated by it, but

1:08:60 you know, it's funny because I think through like what should the last question be and it's always like what's the controversial take that you'll die on a hill about And this is it. And I think the

1:09:09 beauty of it is to your point and to what you look for, Akila, which is someone who is going to identify the problem and has a bias for action to execute it and will own that entire, you know,

1:09:19 structure is what you're doing now with this. With what an AI native company organizational design is going to look like. Right? It will change. It is a problem how it was built before. And

1:09:32 here's how we're doing it doing it. Yes, I was doing. I think, again, if there is a takeaway, it's like, We can build anything right. That's that's the world. We live in today We don't need

1:09:41 30 40 engineers to build stuff. We need three or four right that took 30 40 people to build three of our engineers can do today so then it's about Making sure that we're prioritizing the right things

1:09:52 to build again going back to my journey It's like the right is this the right thing to build and more importantly what not to build because we can build so much Right, it's like yeah, I can keep

1:10:01 adding features, but that's not the point right. That's only gonna overwhelm the end user and keeping the end user in the center of user design, right? It's like Clearly knowing who they are what

1:10:15 their strengths are and what their weaknesses are, right? Oftentimes I think startup entrepreneurs live in a bubble, right? They're drawing from their own experience I think this is an important

1:10:25 problem to solve which is great but if you're not thinking about who your end user is and if you've not spent enough time with them and understand them to the core, right? What motivates them?

1:10:36 Because. I believe that humans hate doing work, right? Including me, right? We don't like to do work. Especially if it feels like work. If it feels like work

1:10:47 and if you're asking me to do more work to get certain return, I think it's then what I'll do is I'll start calculating is it worth that return, right? But then if you can take away that work

1:10:57 aspect completely off of my plate and if it feels that way, right? Then I'm willing to put in like some sort of an investment, whether that's money or like time initially. And that's how we design

1:11:08 Pi. It's very much set it and forget it, right? You set it up and then it keeps giving you the rewards. And as the data gets better, it's only gonna get you better rewards, right? So the more

1:11:19 data we have, the better we will be. So that's it. And I think that's gonna be the philosophy here. That's exactly it. That's how we grow your slice of Pi. And we wanna, so any new product that

1:11:30 we build, we put it through that scrutiny Is it adding more work to our users? If it is, is this the simplest way we can solve the problem? So let's go back if it's not, right? So we make, like

1:11:41 I said, we're making mistakes on design. That's our bottleneck. So engineering is no longer the bottleneck, right? It's a very different world that we're living in. It's that design is the

1:11:50 bottleneck, which is the great place to be, right? So we are building things that are really valuable because I've had this conversation with some investors. They're like, why are you not hiring

1:11:59 more engineers, right? I'm like, I don't think we should. 'Cause then what happens is the product gets diluted really quickly So many people who can run, we already seeing these people be

1:12:08 extremely efficient. So I'll end up with a product that's like very disjointed user experience because we're doing it for the sake of doing it, right? We have these engineers, we have to fill

1:12:17 their time, let me go make them do something. Or they're cleaning up tech debt on other things. Doing low value work. Yeah, low value work. And I think tech debt is, right now I don't care for

1:12:28 it. We probably have quite a lot of it, but I don't care about it yet, right? I think it'll, what's gonna happen is, or the course of the next tier. Which is, I got to imagine, that's a

1:12:37 different mentality than - That is a different mentality. That's a very startup mentality. Because I even, even with the code, I don't expect our engineers to write the most perfect code. I'm

1:12:46 sure a lot of CTOs will hate me for saying it. It's not about that. It's about, let's go solve this problem very quickly and then see if this is a problem worth solving, both from a money

1:12:56 perspective and a value perspective, right? And then I can guarantee you this and you'll probably hear this from a lot of CTOs. They'll go through a cycle and then they have to go re-architect the

1:13:05 code, like just that happens with any company, right? So you know that that's gonna happen. So rather be quick and get it out there, get some feedback and then build on top of it. Versus, oh,

1:13:16 let me make this perfect today. The perfection comes in the architecture, the modular piece, not like the nitty gritty detail, right? That's the distinction, if that makes sense. It's like the

1:13:27 perfection is, okay, here's the modular stuff, I have an eye for the vision, right? but not so much, okay, I'm am I writing this in the most concise with this, like one function in my program,

1:13:37 the most concise way possible, right? That's the distinction. Okay, yeah, I mean, I think this could be a whole nother two hours in

1:13:47 discussion, but - I know, I'm sorry. 'Cause I think I have a gap in my mind of like, how do you build something modularly that isn't as clean and simple as possible, right? Because then when you

1:13:57 do take it and try to apply it somewhere else, imagine that you've got a lot of nodes that could potentially conflict with this new spot that you've plugged in that mode into. Yeah, I'm trying to

1:14:08 think of a good analogy to kind of describe it. So think of it almost like a black box, right? So you know the input and you know the output.

1:14:16 Is it the best way to get that output? That's a black box today. So that's the modular aspect, right? So what I'm saying is have a lot of awareness about the input and output of that box. That

1:14:27 box and how he gets to that output maybe not most efficient way, right? It's not writing the right queries. They're very, they're not performant. They take a long time, blah, blah, blah,

1:14:37 right? That kind of stuff you're not optimizing for today, right? But the input and output, that is quote unquote, set in tone, set in stone, right? I think that's the point in seven. Okay,

1:14:47 okay. So the detail of that box, that can change and we'll go back and likely go change it 'cause we've learned a

1:14:54 little bit more. But it's the modularity of the loop. Of the loop, yeah. Got it, okay And that's kind of set and that also gives you a framework for when do you split it into a separate module,

1:15:05 whilst skipping it in this module, right? Yeah, yeah, yeah. So it kind of forces you to think about it that way. So that's what I mean by having that architecture set, if you will. Akhil,

1:15:15 thank you so much. No, thank you for having me. And this was awesome. I gotta get you to your next meeting to go continue building value. Yes, it's a

1:15:24 fun day today, yeah. And then back to Boston. Yes, back to Boston tonight, yeah, it's awesome Thank you so much for having me, really appreciate it.