Tech Shift Survival Guide with Aparna Sinha, SVP of Product at Vercel

Speaker A: And so we had all started building and I think over time we just got used to, okay, every day there's something new. The eyes are always open, the ears are always to the ground. And I think being here in San Francisco in the middle of things, it helps a lot because you are sort of in the middle of shaping the revolution. That's kind of what Vercel's role is.

Speaker B: This is Next Gen Builders, the show for the growth and product leaders of tomorrow. Cloud, mobile, SaaS. We've seen the big shifts before, but this one, it's deja vu on fast forward. Of course we're talking about AI. AI is moving faster than any technology wave we've ever seen. If you blink, you're already behind. On today's episode, we'll be talking about building in the middle of the AI tornado. How you can stay ahead when everything keeps changing, and what it takes to lead through tech's most chaotic moments. Today we're going to step inside the eye of the storm with someone who's been at the center of multiple tech revolutions. Aparna Sinha is SVP of Product at Vercel. Welcome.

Speaker A: Thank you, Francois. I'm really happy to be here. Very excited uh, to speak with you and um, to talk to other product, marketing and business leaders.

Speaker B: Awesome. Well, let's start to talk about you, uh, let's talk about your career because you haven't had the most traditional career started, I think at Stanford and now you're the SVP of Product at Vercell. Walk us through your journey. How do you get to where you are today?

Speaker A: Yeah, you know, I don't know that there is a traditional career, especially for product managers. Product managers and product leaders come from all kinds of different paths to product management. But uh, I think something that we share in common is a real love for users and for customers and ability to identify patterns and see what uh, customer needs are, um, as well as like a business acumen. Uh, and so yeah, along the path of my career I discovered that those are the things that I was most passionate about was um, really strategy, understanding customer segments. Um, I'm really, really deeply driven by technology, particularly cutting edge technology. You know, I tend to uh, run towards the latest, um, deep, uh, technical changes, um, not sort of like small ones, uh, especially high risk ones. So that's just um, I guess how I came to this career. I started um, at Stanford as an undergrad studying physics, um, which uh, maybe it's an unusual thing but I like to understand how things work, particularly uh, in science. And engineering. And I found uh, during my studies that actually the most uh, interesting part was um, how computers work. And so I got involved and I got interested in computer architecture and realized that I guess it's not exactly physics. I moved to electrical engineering. I don't really care what the titles of these things are. But uh, so uh, I did my PhD in electrical engineering. I like building things. I actually really like building things, uh, with my hands and uh, creating things for people, things that other people will use. Which I guess is the definition of an engineer. Absolutely.

Speaker B: Um, and is that how you ended up eventually going to NetApp?

Speaker A: Yeah, eventually. Uh, well actually after my PhD, um, you know, I worked uh, as a research, uh, staff, but member of technical staff briefly uh, at intel and briefly at Agilent. Uh, uh, those were very brief stints. But then I went to McKinsey. McKinsey at the time was hiring what they called advanced PhD candidates or advanced degree candidates into these technical disciplines in their high tech practice. Um, so that's how my career started. I again wanted to understand, uh, not just how products are built, but how do the products, how do they get shaped, um, and how do they make it into uh, the hands of the user and which ones become successful. So that's just kind of going from building things myself to wanting to uh, build something useful for other people. Realized during my time at Stanford that that is what I wanted to devote my life to, building things that are useful for other people. So I went to McKinsey to kind of really learn how big companies do that and how do they get these products into the hands of people. Five years at McKinsey, I learned a lot, and I think particularly about myself, that I love working with customers. I am an abstract thinker. I like to connect the dots, see the patterns, what are the uses. Um, and McKinsey in general, you want to be few years ahead of the game. You know, you're, you're doing strategy, you're looking out several years ahead of the game. So at McKinsey, myself and a few others, um, we started and became part of the cloud computing practice, which at the time was a very emerging area. Most of our clients which were large enterprises, were not going to any kind of cloud. They had servers, they had data centers, and no one was going to go to the cloud. And so we started this thing called the cloud computing practice, which is like, not only will your software be in the cloud, but your servers will be in the cloud. Your entire everything infrastructure will be in the cloud. And, and I kind of focused on this interesting layer called platform as a service or paas, which is that you don't need to be you, the user, the customer will not need to be aware of the servers. You will focus on doing your work which will be whatever it is. Maybe you're a developer and you're trying to deploy an app. So that became my platform, if you will. Platform as a service became my platform as a consultant, um, at McKinsey.

Speaker B: Are there some frameworks that you learned at McKinsey that you're still applying today or you're in this transition phase where the cloud was being born and now we're in this new phase where AI is being born. Are the lessons of then, of how you approach problems, how thinking about problems still relevant today and still something that guides you today?

Speaker A: Yeah. And you know Francois, many of your viewers probably have an mba. I have no mba. I don't have any business training actually I have a PhD in engineering. And so McKinsey was my training ground for it's I, we call it like an express or working mba.

Speaker B: Ah.

Speaker A: You know, I learned all my corporate finance, M and A. I also strategy, marketing, everything having to do with how to run a company, um, and how to uh, execute a strategy, a long term strategy for a company or even how to create one, how to get alignment and buy in. And most importantly I think the thing that I learned at McKinsey, which obviously builds on my own talent or my own natural kind of tendencies, is understanding uh, the market, understanding customers, analyzing, finding those trends. Um, and then the analyzing part is obviously very mathematical. And McKinsey, typically people create uh, these models of the future, um, with sensitivities. And you can do that for pricing, you can do it for um, market entry, for growth. Um, there's so many different ways in which that applies. Uh, it also applies to infrastructure. Like if you're analyzing a trend like movement to cloud there, you know, you'll generally analyze the supply chain so you'll think about who are the providers and who are the consumers. What is the economics for the consumer that would make them move to this? What are the forces that work for the supplier? How will they need to change both their offering as well as their pricing and packaging as well as their cost structure in order to go into this new industry structure. And we're of course now in the midst of a similar change, uh, with AI where the industry structure is changing and the value, uh, structure is changing and the pricing structure is changing and how, you know, where's the value to the user, who are the users? What are they going to consume and who are going to be the players? Uh, we're seeing it sort of like play out. And I would say every shift is different, but there are certainly things that you can learn and there's certainly a toolkit that can be useful.

Speaker B: That's fascinating. And, um, when you were in those days of trying to get customers to move to the cloud and deal with an uncertain market or unknown market, they had to take a leap and a leap of faith, I think, about those lessons, especially today, about AI. And we're building an unknown future, which has a lot of possibilities but unclear outcomes yet. You know, how do you drive that change with clients? Right, to adopt these new technologies, to adopt a new way of thinking, a new way of operating.

Speaker A: Yeah. Um, and for me, because I am a technology leader and my passion is, um, in the technology and that goes back to sort of like, you know, just my roots, it often starts with understanding the technology first. And so, um, uh, that involves, of course, playing around with the technology and understanding what is the fundamental difference, what, uh, is the fundamental breakthrough and then where is this going to go? Typically, when you're in the early stages of a technological change, um, you want to assess, like, how big of a shift is this, how long is this going to take, how real is it? What are the big problems that still need to be solved? What's my prognosis for when these problems will be solved? And why, like, why do I think this will happen? Once you understand for yourself, like, what that innovation is, then it's a lot easier to talk to a customer or potential user about what it could mean for them. And of course that requires having sat in their shoes. Um, and I think my, my early career at McKinsey and then also my most recent work at Capital One, which I would consider sitting in the, in the users or customers shoes. I think it gives me a good sense for the types of stresses, pressures, uh, uh, constraints, uh, and goals that our customers have. So that's really important. Uh, I think in any kind of business role is, uh, people call it user empathy, but it's really knowledge of, um, the environment of your users and what are their goals, what are they really trying to accomplish. So then you marry that technical breakthrough and, and what it could be before it is, you know, to what it could be to the needs of the user.

Speaker B: That makes sense. And I love, you know, the expression of, you know, being in a, uh, customer's shoes. And it's one thing to say it It's a different thing to live it. And you've lived it. Yeah. And you've also lived different experiences. I know we jumped from McKinsey to Vercel, but in between, you've had NetApp, you had Google Capital One. Like, tell us maybe a little bit of like, why you made those changes in your career. What were you trying to learn? Experience. And what are some of the big learnings you had, uh, through those different, uh, experiences.

Speaker A: Yeah. So I knew that I didn't want to be an advisor. That, um, my personality is very much a doer, builder, seller, you know, all of those things. An operator. I would call myself an operator. Uh, so after five years at McKinsey, you know, I felt like I had gotten all the tooling, um, as well as a good network, as well as, like some sense of, you know, business judgment and all of that. And I knew at that point that I wanted to do product because I love technology and I love customers and um, I like to build. So it was just the right role for me. And I had never done product management. So people are like, well, do you want to run corporate strategy? I was, no, I do not want to run strategy. I want to. I want to build and ship product and I want to sell it and I want to make a lot of money for the company. Like, that is what I'm driven by. I'm driven by, uh, you know, P and L ownership and user satisfaction, uh, and getting value to customers. So that's, um. It became very clear to me at that time. And yeah, I went to NetApp because it was just, you know, it was, uh, actually a good company at the time. They were working on private cloud. They wanted me to lead the private. Private cloud manageability, private cloud software. And then I went to. Pretty quickly went to Google from there. Google was interesting because they didn't have, or they had a very nascent enterprise offering. And I had mostly worked in enterprise. I went to Google also because I didn't want to only do enterprise. I wanted to try consumer. Uh, and Google was a very good fit for me because it had, um, that, that deeply technical, probably even more so than NetApp. Deeply technical, um, product in engineering culture, which was a super, super good fit for me. Um, and so I would, I stayed there for ten years. Uh, I definitely really fit in, uh, enjoyed. Uh, I would say I'm a Googler at heart now a zoogler. Um, but yeah, and I've, uh, worked in different parts of Google. I started in the consumer group, uh, actually on Android which is an operating system still a, I'm still a deeply loyal Android user. Um, and that is just the whole sort of operating system area is an area that ah, I feel aligned to. Uh, it's kind of related to my PhD, you know, how computers work. After a couple of years at Google I found this other open source project which was more enterprise oriented or at the time it was just developer oriented, uh, called Kubernetes, uh, which is also open source. It was a very small team and a very technical team. Entirely engineering but very technical. Most product managers at Google who are maybe consumer product managers would not want to join the Kubernetes team at that time.

Speaker B: Right.

Speaker A: It's like 2015, 2016, um, would not want to join the Kubernetes team because so hard to pronounce what is this thing.

Speaker B: And Kubernetes at the time was purely open source. There was no commercial model yet it

Speaker A: was purely open source. Um, we had started hosting it on gce which is the VM based offering more for test purposes, more to see like okay, this thing that we're putting out in the world, does it also work? But not really a commercial angle, not really with the goal of building a business. And so it was the perfect opportunity for me an operating system not for your mobile phone but for the cloud, for everybody, for all enterprises. I was initially the open source product manager which is like open source product manager. Um, but yeah, um, I did help to set the open source roadmap, um, working with many of the community members. I certainly gained from working in an open, uh, open community, a global community. Uh but my contribution was really to build Google Kubernetes engine which is a hosted service of Kubernetes on Google Cloud and focus like have a focused team both from engineering. I built the product team, uh, built a business, became the P and L owner, um, built that service ah, into a multi billion dollar service, the third largest revenue generator for Google Cloud, um, over, over several years. Yeah. And um, it was a fantastic experience for me, growth experience for me.

Speaker B: So let's talk about this transition to Vercel and your role there and maybe stepping back. Can you explain what is Vercel? And some people may have heard of V0 but not really know what V0 is. Uh, give us, give us the overview, give us the pitch of Vercel.

Speaker A: Yeah, so Vercel, I first heard of Vercel when I was at Google. Um, because uh, if you remember talking about thinking about platform as a service or paas with the idea is that you can just build your application and the infrastructure will just take care of it and you can host it at scale. That's what Vercel is. Vercel is basically framework defined infrastructure which is that you have an application, you define what you want the application to do. And um, we uh, you know, you see this NEXT logo behind me. Uh, Vercel is the creator of next, which is an open source framework for web applications. It's one of the most popular. Um, it's a standard for web applications that was pioneered by Vercel. Um, it is fully open source, similar to Kubernetes, has an open source community. Um, but a lot of the maintainers are here at Vercell. Um, and the infrastructure that Vercel hosts is in service of this framework and other frameworks. So Vercel is now a much more general purpose cloud. It uh, is optimized for many different types of web applications, front end and also backend. Um, and over the years Silvercell has been around for I think 10 years, although really um, has been in sort of explosive growth mode for the last five years. Um, with the success of next, uh, and uh, then with Vercel's expansion to cover all frameworks for web applications. And now uh, over the last couple of years Vercel um, has expanded uh, into providing the primitives for creating and hosting AI applications. Starting um, with again an open source project called the AI SDK which is an open source SDK for developers to create AI applications. AI applications like AI chatbots, um, and also multi agent applications. And Vercel excels at easy developer experience going back to you can just imagine an application and you can just build it. And that's also the heritage from which V0 comes. For those of you that have heard of V0, it's a vibe coding solution, um, also created at Vercel and hosted on this cloud that we're talking about.

Speaker B: Yeah, uh, it's an amazing solution and the unlock that it provides in terms of reducing complexity but also empowering creativity is just incredible. You're building AI solutions, you're building these new frameworks, uh, but you're also building on an environment where it's constantly changing, models are changing, expectations are changing, the art of the possible seems to be changing every six weeks with new capabilities that are coming up. Um, how do you build in a rapidly changing environment like that?

Speaker A: Yeah, and I think that's why you need to partner with a company like Vercel or, and uh, certainly why I am at Vercel actually when Vercel was founded, I think it was called uh, Zeit, uh, and then maybe now uh, because it's a company that's very much in touch with the Zeitgeist or with the current, with what is now, what is coming and what is now. Vercel is also sort of a hotbed of activity for startups, um, both, um, literally physically. You know there's a lot of startups that um, are in our office day in and day out. Um, but also uh, technically there's a lot of startups that are hosted on uh, Vercel, um, and that start their um, life on vercel. Like with v06 and a half apps are created every second.

Speaker B: With Vzero every second.

Speaker A: Every second.

Speaker B: That's unbelievable.

Speaker A: So um, yeah and that is a pace that I don't think really existed before Vibe coding. And now you don't have to be a developer, you don't have to be on GitHub at all.

Speaker B: That's great. So when you're as a builder and as a leader of teams, does that force you to rethink how you build? Like can you even do 12 month roadmaps? Can you even uh, plan when everything is moving so fast all around you? Like how do you manage right, your teams and innovation in that pipeline?

Speaker A: Yeah, there are multiple loops. I mean and actually I don't think it's any different than even if I think back to like my career at Ah, McKinsey, you know you're typically trying to make very, very long term changes but you are doing it every day. So you're doing small things, you know, every day and it is in an environment that's constantly changing. So you are actually responding to the environment as well and you're very much like aware of what's going on. And I think in AI that's, I certainly have gotten accustomed to it over the last two years. So I left Google in 2022 uh and I went to a startup accelerator. I started an AI accelerator at Paravisi. Um, and I started working with like um, uh 15 different teams of um, founders, like straight out of college founders, you know and they had no notion of like 12 year, 12 month roadmap or plan or like large teams. You know, you have like teams of two people that would just like overnight build something, the next day build something else. And so my whole um, everything got reset once I ah, and it was an AI accelerator at the time when I think um, GPT 3.5 if you can, if you, if you remember that. So Long ago.

Speaker B: That was like, it's an eternity ago now.

Speaker A: Yeah, end of 2022 that's when that had come out. And so we had all building and I think over time we just got used to okay, every day there's something new. Um, and we just uh, the, the eyes are always open, the ears are always to the ground. And I think being here in San Francisco in the middle of things, working uh, with all the startups, it's where I want to be. Um, and I think it's, it helps a lot. Um, because you, you know, you are sort of in the middle of shaping the revolution. That's kind of what Vercel's role is.

Speaker B: That's great. So I mean you brought up uh, GPT 3.5, which wasn't that long ago, but feels like an eternity in this world. Uh, just recently as we're recording this, GPT5 just came out. What's your view on that? Like how does it change what you build? Does it unlock new possibilities or does it make you rethink assumptions of what you've already done?

Speaker A: Yeah, well I mean along the way there were a lot of other things. Um, there were a number of uh, innovations in image models and um, voice, um, and uh, speech to text. I mean all of that has gotten very good. Um, but also in code generation which is perhaps the area of most maximum interest to me given my background in the developer space at Google and the work that I do here at ah, Vercel. You know Anthropic has had just the leading model in this space for um, at least the last year, um, with Sonnet, um, Cloudsonnet, uh, 3.5, um, and uh, that has just been an enabler for so much of the ecosystem. And that I think is what has led to sort of this huge revolution in how engineering is done and how AI assists with engineering. Everything from uh, correcting code to writing new code to planning new code to um, reviewing code, um, to um, any number of engineering tasks. Uh, I think that that is the first thing to mention is, and that is also what led to all of the vibe coding and so on. Um, there are other um, models. Um, there's MOD from Gemini. Gemini Code Assist has become really good at code generation and so you see it like every few months. There's a model that sort of leapfrogs the previous model. I think it's great. This amazing innovation at Vercel we have um, the AI cloud. Uh, a big part of that is um, our AI gateway. So our AI gateway is this Service that allows um, uh, users to connect to over 100 models including of course the um, Cloudsonic models. The Gemini. Gemini has a whole family of many different models, Pro and Mini and Nano. There's so many different models the same with OpenAI. Uh, and so the gateway is a way to connect to any of these models. And in fact one of the value propositions of the gateway is that allows you to connect to the same model from different providers so that you can have a consistent level of service and you can sort of switch between providers um, to maintain that consistent level of service. And because we are the creators of V0 we ourselves use these models, you know, uh, Claude Gemini, uh, OpenAI. We use these models in a vibe coding context which is it, like I said, six and a half applications per second means it's very scaled usage and so we have good um, volume with these providers, with the model providers which um, uh, you know gives us access to a higher rate limit, um, which we are able to provide to our users through the gateway. So the higher SLA through balancing across providers and then the higher rate limits with some of these uh, Frontier models, GPT5 in particular, um, there are two things that happened. Was it last week or the week before? I can't keep track but um, we're

Speaker B: in a time warp.

Speaker A: Yeah. One of the big things that happened was that OpenAI open sourced, um, you know, state of the art, cutting edge models, two of them, ah, 20 billion parameter and a 120 billion parameter and I think now it's just like. Oh yeah, that also happened but that was like a big deal um, because uh, you know they had not open sourced um, and they are one of the leading, if not the leading provider of these, of these models. So I think that's a big boon for many enterprises, um, you know, who want to have a model that they host themselves. And of course you know we made that available in our gateway within I think we were already working. OpenAI provided us a little bit of early access there and we had several providers um, that uh, you know, that were part of the gateway. Uh so we were able to launch that and then a few days or I think two days later GPT5, the three GPT5 models came out and we had, we were a launch partner with OpenAI. Um, so Vercel was a launch partner partly because of our expertise with next. So um, when you are that involved with a framework, these models they want to be optimized for a framework like next. And so we've developed evals for um, Next JS and also for AI SDK. So we're able to provide feedback to the model providers, any and all of them, whoever wants it, um, on you know, how well the model performs against these evals.

Speaker B: That's great. Looking back at ah, your career, you know, I love asking the oh shit question or the oh shit moment in your career. Uh, what's a, what's an example of one of your experiences where you had this oh shit moment?

Speaker A: Well I think nowadays it's pretty frequent, you know, because um, I would say that it's in a positive way, um, really good, um, when you work with good engineers in a um, highly ah, productive culture. Um, which is what I would say we have at Vercel. Um, we've got a small team relative to someone like a Google or um, even some of the AI labs. Uh, we're about 600 people growing very fast. Um, but I would consider that still a smaller team where everyone knows each other. Um, and we have this culture that is um, about shipping. So if you ever see like kind of the logos, uh, uh, and the signs on the 101 about Vercel, um, if it's fast, it must be on Vercel and you can just ship things. Um, that is not just a logo. That is the way that we live. And I think um, one of the key moments for me um, in joining the company, uh, was realizing that you can just ship things and uh, that you can work extremely fast, um, and that you can partner with really large companies like OpenAI and Anthropic and you can enable your users to have access to this technology and that you can actually translate your um, expertise. In this case it's in uh, developer experience and it's in the web. You can actually parlay that into the um, next frontier of um, what AI can do for the web. Uh, for me that's a huge, I um, think moment of change and moment of excitement. Um, and so I think we're uh, living it right now. And the biggest sort of um, uh, key moments these days are in being able to enable some of those dreams. A lot of things, at least in my career are coming together. Um, you know, when I first started in open source, maybe one moment for me was like, oh my God, open source. How am I going to monetize this and is this going to be possible? Or maybe I will not be able to monetize. Maybe we will not be able to maybe. And if we can't then it won't be a sustainable business. But then we did. And you know, um, I think um, I did certainly contribute to like how did, how to monetize something that's open source and still keep the soul, still keep the community, still keep it open source. And so now I feel like you could certainly do that in AI. I think I maybe have learned from that a little bit and feel like that that's doable. Same thing with um, paas. It's just been a long journey understanding um, uh, how to get to that point where you make the developer experience really delightful. And now we're here, I don't know, maybe it's nirvana. Now we have vibe coding. Another, another thing is like, but if we have vibe coding and this is, this is the new thing, this is the new. Actually maybe another moment is like, oh my God, how do we make this secure? Wow, there's AI. And AI can be really insecure. Um, and that's the benefit of I uh, think working with a world class team here, a small team, but a world class, tight knit team that has this mentality of ship and make secure is um, some of the work that we've done on security I'm really, really excited about. Meaning, um, uh, where there are a lot more threats on the web now because of AI, there's a lot more uh, bots on the web and there's a lot more agents on the web. And so one of the things we're doing is just recognizing what is an agent and recognizing what is a bot and what type of bot is it. Is this a bot that's scraping your website for a particular purpose, like for example to uh, allow ChatGPT or Claude to provide answers? We, or is this a bot that is um, scraping your site, um, in order to be able to sell your wares? And again maybe this is. I love working with customers. We went back, we talked about that. That goes back forever. I'm learning from all of the customers about what they want. They want certain bots, they don't want certain bots. And so we're now creating a bot protection service, um, bot management service that enables you to know which bots are coming to you and which ones you want, which ones you don't want, which I think is critical and also recognizing agents. And there are good agen agents and agents that you may not want. A big question is that if your customers are coming and they're talking to your agent and your agent is using something expensive like OpenAI tokens or cloud tokens, well, if you get hit by a bot that's Then using your agent, that could really run up your bill. So how do we protect our customers from those kinds of bots? And we've got a new technology, new capability called bot id, uh, that is able to recognize what's a human and what's not a human, which is amazing. How to recognize what's a human, what's not a human. It's very sophisticated. It's a lot of machine learning. And so I'm sort of like in this rebirth, maybe phase of my life where, like, all this technology, new technology is needed to make the new technology safe and possible. And it's like, um, I feel like a child again, uh, you know, being able to, to be part of this and study it and understand it and again, like, find where it could fit with customer needs.

Speaker B: I love it. I mean, you hear it in your voice, the passion, the excitement. You're making me dream and you're making me, uh, inspired, which is fantastic.

Speaker A: It's a good time.

Speaker B: Absolutely.

Speaker A: It's a good time for everyone.

Speaker B: So just as we wrap up, what is one piece of advice you can give to aspiring builders on how to succeed in the AI era?

Speaker A: One piece of advice. Um, there are. There is so much, uh, new enabling technology that is so accessible in the AI era. I, um, think it is important for builders to use this technology and also find ways of building with it. And I think it has never been easier to be a builder because you can literally have the AI teach you how to build in whatever language or whatever format you may want. Beyond that, I would just say follow your heart and follow your curiosity. Whatever speaks to you, that's probably the right direction to pursue.

Speaker B: Well, that is great advice. And Aparna, thank you so much for joining us. Uh, and thank you all for listening to next gen builders. And look out for our next episode, wherever you get your podcasts. And please don't forget to subscribe.