Tyler Postle, Voker.ai: Co-Founder @ Voker.ai (YC S24)

Welcome to another Founder Wisdom podcast. We have Tyler Bosel with us today. He's co-founder at uh Vulker.ai, which is a Y Combinator summer 2024 company.

Uh Tyler is a pretty smart guy. He was a CPO at a bunch of companies. Uh and yeah, started that one about 11 months ago.

Uh Vulker Verer is an interesting company. certainly leverage uh AI uh to help people create better products. Uh I'll share my screen in a couple of moments, but uh Tyler, can you tell us a bit more about yourself and about why you decided to actually start Vulcer?

Yeah, thanks Charles. Yeah, I appreciate you having me on. Um yeah, a little bit about myself.

My background is in in data science. Um I actually started out as a physics major in in undergrad but caught the entrepreneurial bug early on and uh I joined a e-commerce startup as one of the first analysts built out the data science function there and um have been interested in machine learning and AI ever since. Very cool.

And you said you caught the entrepreneurial bug. Why was that? Because I see you've also worked at a couple of companies.

What was your your first catch of that bug? When was it? Yeah, it was actually in college.

I went to UC Santa Barbara um and uh there was an entrepreneurs association on campus and they were hosting this event called Startup Weekend which is uh sort of organized and sponsored by Tech Stars and um I decided to attend and uh really just love that rush of you know working hard for a weekend, you know, pretty much barely sleeping and and just working the whole time and coming up with an idea and um going out and validating your product in the market. And so that's kind of where I first realized maybe I wouldn't uh continue on the research journey um and uh started looking into sort of early startup opportunities coming out of college. Hackathon styled and what are the parallels uh between physics and the startup world like do you think in physics when it comes to building a startup and a product?

Yeah, definitely. I think um it's kind of hard for physics uh physics degree to not shape the way you think about like the world and the way you approach problems. I think there's a lot of problem solving obviously when you're starting a company.

Um and so you know you hear this all the time. People talk about first principles. Um one of the things that I think about a lot is sort of the the extremes in a lot of physics um phenomena.

you kind of have like maybe the minimum or the maximum case um when you take something to infinity or to zero. And so lots of times when we're thinking about solving a business problem that framework um comes to mind for me just from my physics background. And how is that is it um yeah how reality works?

Uh is it getting from zero to one laws of speed momentum terminal dynamics like how do you tend to think in terms of business? Yeah. Um, I think probably a lot in in kind of equations, it's kind of like what are the inputs?

I mean, especially when you're building um sort of a an operating motion, whether that's go to market and how much money you want to spend on on ads or how much money you want to invest in building out a team. Um, it's kind of thinking like what is the sort of the graph look like over time of okay, if we put this much time and energy or money into hiring out a team, when do we get the returns? Um, those kind of things.

and uh kind of forecasting out in this sort of mathematical equation cuz I saw yeah that you were passionate about building high performance teams. Um in your opinion, how did uh open AAI happen? How did they reunite so many talented individuals in in one room?

Have you analyzed that one? You know, I haven't I I haven't like spent a lot of time looking into that, but I I could give a guess. And my my guess is normally I think you have to be pretty focused in business in general.

Um especially in the startup world, right? Like there's so many things you could do, so many decisions you could make. And so being really clear on your vision, whether that's about where the market's going to be or how you should build a company or a team.

I think that having like it's kind of like with marketing messaging, right? You want to find the right audience to talk to and and share the right message to them. It's the same with building a team is having like a clear and consistent message and realizing it's not going to resonate with like 99.999% of people.

Um, and that's okay because you want this like very focused team that fits well with um, you know, your skill sets and the way that you see the world and um, yeah. So, how did you pick your co-founder? Let's start with that.

Yeah. Um, Alex and I actually worked together back at uh the first uh entrepreneurial company I worked at, um, which was LOL.com, a a bootstrapped direct to consumer uh, mattress brand. And Alex was an early uh, intern on on one of our data teams.

Um, and at the time I had a a director of data engineering who was kind of leading that team. And so I didn't even work with Alex directly that closely, but I was always hearing these updates of like, you know, oh, Alex just spent the weekend rewriting the whole codebase and it's like so much better than anything I could have done and different things like that. And so I kind of knew um we had someone special there and and kept an eye out on him.

And then over time we had the opportunity to work together in some consulting um gigs where um I had some of the strategic connections and um kind of was helping with data science and analytics strategy and Alex was the engineer in the background building the the scalable systems. And so um we just loved working together and it was a good um sort of pair of our skills and so um over time we were continuously brainstorming ideas for for a business and that ended up with with Boker. Tell me a bit more about uh LOL.

What was the engineering behind that one? Yeah, so LOL was pretty unique for uh for a direct to consumer company. You'd think that, you know, um maybe not super focused on the engineering or data side, but that's actually not true.

Um a large portion of our team was devoted to data science, data engineering, engineering in general. We had uh the one of the founders and the CTO of the business had actually developed his own tracking system like back before um UTM parameters were were super popular. Um and so we had an internal tracking system that let us track like every touch point of a user before that became super common.

People were still using like last click and we were already thinking about okay multi-touch you know um incrementality. And so because we were bootstrapped, we really used data science and our marketing engine to really make sure that every every dollar we were spending was actually incrementally profitable rather than just like having some large budget and spending it on billboards or something you can't track. And so we had a large data function to do the predictions around how likely is it for this ad to convert?

Um what's the incrementality of that dollar that we spent? Should we be spending more or less? And um my boss, the the president and CMO would always say like, "Hey, well, we have an infinite budget as long as it's profitable." Um and so that was kind of the the pro the math equation we were trying to solve there.

That's that's always easy to say, right? That getting profitable is incredibly hard as you probably know. Tell me about your YC story.

How did you get in and how was that game changing for you? Yeah. Um yeah, like I said, Alex and I had been um doing consulting together.

we both were working kind of in the data space. Um, and we had been following YC and familiar with it. Um, but you know, it's it's kind of tough to make that leap to start your own company.

Um, really want to be all in and be able to to spend the time and have the um, complete focus to to build a business. And so, um, you know, Alex had experienced some pains at the the job he was working at with the state of the tooling that existed for building with LLMs or generative AI. And so, he had this idea that he came to me on one of the weekends that we worked together um on our side projects and he was like, "Hey, I think we should apply to Y Combinator with this." And I thought it was a great idea.

And so, um, we jumped in. In terms of how we got in, um I think we got, you know, pretty lucky because it is a super desirable and a tough tough environment to to join, but we're so grateful because it really um really changed the trajectory of our business and um helped teach us so much in such a condensed time. Um so, and which are your favorite uh PG program principles?

Oh, good question. I don't even know. I think uh I I honestly lean more on the maybe I'm biased because Dalton was our group partner, but I lean more on the the Dalton and Michael videos um on YouTube.

They're they're a great pair. We got the the opportunity to work together with them as they were our sort of mentors and group partners throughout the batch. And uh it's a good kind of balance.

Yin and Yang, they both have great insights and Dalton's a bit more reserved and and methodical and Michael's definitely provocative and uh and outspoken. And so um I don't know I just think a lot about those those uh kind of conversations that that they shared with us right and you seem to be at the right place at the right time right like shipping AI features in minutes no coding required guess a bunch of people are uh vibe coding nowadays but they're not like fully completing uh their app into a working prototype uh tell me about the vulkar idea and why you suspect uh we may get product market fit. Yeah, definitely.

So, yeah, there's a lot of companies that are doing great things focused on like this early prototyping or even even writing code, but obviously AI is not yet at a place where it's building itself. And so, you still need um you know, good teams, software engineers, AI engineers, product managers to work together to build actual AI features. So, Vocer helps you build LLM powered features into your software, into your back office, into your product.

And so if you aren't a large company like Microsoft or someone who can invest tons in an in-house team or infrastructure, you could just use Vocer's platform. We have a UI that makes it easy to prototype and build LLM features. Um it lets you share them across stakeholders.

So a product manager might have an idea for how AI could improve a workflow in their own software. They can prototype it themselves without bogging down their engineers. They can then show an example to a stakeholder, get some approval before actually devoting the engineering resources and all the engineers need to do is use the API endpoint for that specific prototype they built on our platform and just integrate a front end around it in their in their software.

And so we kind of abstract out a lot of the um the busy work and the redundant like infrastructure that you have to stand up like vector databases and calls to all the different LM providers make it easy. And compared to your competitors, where are you going that they aren't? Yeah.

Um you know there's the space is really crowded. There's a ton of great tools being built. Um, we're really focusing on those businesses where AI can be transformative for saving a ton of money for their bottom line or even improving their consumer experience, but it's not like their core core focus, right?

So, they are not going to invest in building out a whole AI engineering team. It's going to take way too much time. It's going to cost hundreds of thousands of dollars to hire these AI engineers who are rightfully so getting getting paid a lot these days.

And so, um, we really focus on kind of having a low floor and a high ceiling. So we want to make it easy for a stakeholder to preview an example of an AI feature. We want to make it easy for a PM who is technical but not a coder necessarily to build a prototype.

And then we want to make it easier for the developer to integrate these things into their existing systems. Um so you know super hardcore deep um training your own model use cases like that's not for us. Um, but if you want to integrate AI in lots of different places of your business and have actual tangible ROI, um, we want to make it really easy for you to test those things and get them into production quickly.

And who should go super hardcore? And yes, who shouldn't? I mean, first of all, um businesses that are that have a lot of like um just internal data that uh other companies don't and and probably need to have really big budgets because to train your own model or to really fine-tune it well, um you're going to need a great team.

You're going to need a lot of uh infrastructure, but that's a big investment to decide. I mean, that could be millions of dollars. Um, if instead you are a really like bottom line and growth driven business, you want to get tangible results in the next like weeks.

um you don't want to wait years to invest in like making this bet that you're going to have the best I don't know um uh construction data possible so that you're going to be the best construction focused LM and you just want to improve the the manual tasks that your teams are wasting time on or improve the experience in your workflows of your own software for your customers then you should use VER. Interesting. And yeah, where are you seeing most big uh companies stand into that?

They did they create their own uh AIT techchnical departments, let's say a Fortune 1000. Do they already have that? Are they investing in GPUs and creating their own LLMs or are they just not catching up?

Yeah, I think a lot of them are investing in that. I mean, I'm surprised with how many I'd say maybe not even 1,000, but even definitely not the mom and pop shops, but there's a lot of uh, you know, mid-size companies that are like hiring heads of AI and all these things, which is great to see people leaning in um, and investing in the space, but it also begs the question of like how much have you already validated uh, before doing that or how much of it just is a bet because you were told to do it. And so again, I came from like a a bootstrap business where everything we did was an AB test.

We want to measure profitability and success. And so that's kind of how we've tried to set up this platform is make it really easy. It should take you days to get your first AI use case that's going to save you probably hundreds of thousands of dollars.

Um, and then you can move on and continue to like reinvest those savings in the business. Um, if you're a large company with a huge budget, of course, you can kind of make that big bet and uh extend millions of dollars over a multi-year time horizon and hope for a return. So the simpler solution would be to buy Vulker and have my existing engineers for example use you even if they don't have uh AI coding experiences.

Exactly. Like the the integration will be super familiar because it's just an API. Um this is something traditional software engineers are familiar with and they can learn on the platform like okay what are some of the nuances to dealing with a non-deterministic system but we have the monitoring tools we have the the AB comparison tools and things like that in the platform to help make it easier versus like if you do this inhouse uh you need to build all these tools internally and why is everyone going to kind of create the same infrastructure um across all these businesses.

Do some of them also hire fractionals or consultants for example and these could be the middlemen. Uh they know the company's needs. They've signed NDAs but they they still don't know how to make it technically happen.

Could these also be decision makers for you? Yeah, I do see um there's a lot of IT consulting, dev shops, agencies that are doing great work of helping to be these outsourced teams for companies. I think the one struggle though is that like they then have to build again a homegrown system for every single business.

Um it's not their core competency to build like their own internal software that's robust for production use. And so instead again it's a lot of like one-off projects. Whereas where we've seen some of these companies have success is they partner with Voker.

And so like a dev shop would use Voker as the platform that they're able to deliver to their clients a lot faster. Um and it and it scales with them well. Now let's get into the wilder philosophical part.

Do you feel that the uh AI agent revolution has uh already kickstarted itself? Are we starting seeing jobs being slashed by AIS nowadays? I mean, yeah, I I think it's it's starting, but I think like we're still so early.

I mean, you look at um Google Sheets and and Google Slides and you try to ask it to like make a a PowerPoint slide for you. It's still not impressive to me. like I think the capabilities I know where it can be and it's still not there right and so I think um we're only in the early innings of how um you know the intelligence of the models is imp is improving but we also need to integrate these systems into all of our existing software rethink the UI it's not all going to be a chatbot I think that was one of the like things that kind of slowed down this revolution is that everyone thought that every single AI use case has to be a chatbot and that's actually where Boker started was We we were kind of showing use cases where like you could do an analysis with a button.

Yes, an AI is doing it behind the scenes, but does the user need to know that? Do they need to have a whole conversation back and forth with someone when they could just click a button? And so um so yeah, I think there's still a lot of start stops and starts with kind of the UX and um approach to how we integrate AI.

Why do you think these giants are so slow? Um I think it just kind of comes with like size like the bigger you get. One of the reasons that we wanted to go back to working and building a startup is um because like there's something so dynamic and awesome about working with like a team where you're all aligned in terms of how you work, what you're working on, you're not wasting time on bureaucracy or slide decks or other things like that.

And um that just happens the more people you have, you have a lot more momentum or inertia thinking back to physics, but it's also a lot harder to turn the ship. Um so right the chat interface uh what would be a better interface in your opinion like uh is is that would would it be just like a good experience with buttons? What would be the interface of the future?

Would it even be like voice uh command or or VR? I guess like 30 years from from now would be a neural link or something. But short term what uh what kind of interface would be better than a chat?

I find chats really useful, especially if you ask the right questions, right? But people I I think not that this is a a copout, but I think that it really just depends on the use case, right? Like if you have something where you're going to have a bunch of different types of questions, um you know, a chat's great for that.

If you need follow-up deep dive, and it's going to be a different uh sort of thought process every time, like a different person's going to need different information, then yeah, chat can be great. Um, but if you're, you know, on a real estate listing website and you want to compare 10 properties based on your investment thesis, why should you have to type all that out? Instead, just click a few check boxes and hit run and get an analysis.

Um, and that's, you know, one of the examples that we we show on our website. Um, and so I think it really just depends on like the user experience. Voice is great for some cases, but like if you're in a crowded place, like, you know, are you really going to be talking into your phone?

So I guess all interfaces would be personalized to the user depending on the data that they put in the interactions that they had before. If uh chat is better for me and my type of brain, right? Like so be it.

If it's buttons or voice for example, so be it. So yeah, I like that. Um what are some of your AGI predictions?

Are we on track to get it in the Well, first let's start with the definition, right? What's your uh definition of AGI of super intelligence and when uh should we get it concretely? Good question.

I'm I'm one of those uh people I'm a skeptic in general. I think this again comes just from my research in and physics background where like um you know I think as as I started doing research I learned that like oh even the textbooks that were written are just based on research papers which might not even have been a perfect study of like the right uh number of observations to reach statistical significance. And so I I think um there's a lot more that that computers and and AI is going to do.

But I'm not someone who thinks that in the next like year uh we're going to hit some singularity that everything's going to change. I I still just look at the fact that like people still don't know how to use Google Sheets or um even their email inbox or the remote. And so I think that uh there's a lot that technology can do, but it also never um there's always that initial promise when you think of like something something could do.

uh but it takes a lot longer than we think to actually have that become reality. And so we're really focused here again uh on like what are the tangible things we can do today with AI that will improve a user's experience that will save a business money that will make a business more money and we're less you know like worried about the future of will my phone even matter anymore because I'll be talking into a neural link or something like that right and do you feel that AI uh first do you feel it's uh being democratized right now and second if you use AI. Is it is AI fair?

Basically, does it give people with more money more money and does it give people with less money less money? Does it widen the gap? Yeah, first of all, I do think it has democratized um just like the internet let people have access to information that they didn't before.

Obviously, not everyone has the same access to internet, but um I do think I've seen all sorts of use cases where I'm surprised that like folks that I wouldn't have thought is like tech forward actually are way more capable of using like chatubt to like build I don't know logos for their new business that they want to launch and come up with a business model and get the registration filing that like would have taken them so much time and expertise. And so that's really awesome to see um that democratization happening. On the other side about like fairness and and and model bias and everything, there's absolutely bias in these models.

As someone who did machine learning for many years, um and again coming from like this physics background, all these things are trained on data that already has inherent bias. Um it's great that it seems like companies are trying to think about that, but we know that um there's always this trade-off between innovation and breaking things. And obviously businesses are going to uh kind of pick their lane of how far they want to go with that.

But at the end of the day again, the data that everything's trained on and the way that society works is going to play a part in these models. And um it's an extremely complicated problem to think about how do you actually even detect where that's happening? Um but there are some, you know, great teams and folks that are thinking about that, which is great.

I'm also a skeptic. uh as a skeptic, how do you make your decisions? How how much of science quote unquote, right?

And how much uh personal experience on the other hand, how do you optimize your decisions so that yes, you don't make errors and you maximize good decisions and you're not just these overly skeptic people that cannot even make a decision and are sort of sour about life. Yeah. No, that's a great question.

Honestly, it's something I struggle with. I think that coming from this like super data driven background where I I had experience in a business where we had enough volume working with consumers you get a lot more volume on clicks and um you know like conversions and everything that you can make decisions based on you know data and science and trends and graphs and it's a lot harder especially in an early stage startup to like um kind of have that taste and that uh that gut feeling and so I think number one um is having a team that like helps balance out the things that I you when I get into Yep. Exactly.

Maddie, Vahan, other folks on the team that really balance out um when I'm stuck in like a a loop of over analyzing something, I'll be like, "Maddie, like help me just make sure we we actually get something done." Or, you know, uh Bahan's actually sitting there talking to tons of people when he's doing product research. And so, okay, what do you what's your best assessment from these 20 conversations? Sure, it's not stat sig.

It's not a million conversations, but we can use past past experience to make our best bet at what to build and then we'll see how how our customers use it and then iterate from there. So, do you actually trust your gut or is it mostly data? Because I also struggle with that.

Sometimes I don't trust my gut. You know, my gut is telling me something and it's like no, I need more data before I make a decision here and vice versa. I think it depends on the the category like uh I mean I really do lean on data, you know, as much as I can.

Um that's just kind of how my brain works. Um but there are places where I trust my gut around like um you know investing in the team uh the folks that we bring on board um uh listening to the customer like there's sort of these principles that I trust my gut on but then when you don't have you know 20 million customers how do you decide you know exactly where to go that can be a challenge and so that's where it helps to have a team of folks who are kind of execution driven too to push us to like let's try something and then a week later it's like oh now we actually got that data because we tried something, right? I love um deep researches for that, you know, to get like industrywide data that is fresh and from multiple sources and putting it to various lens of people that I respect, right?

What would Paul Graham do? What would Pat Collison do if if that would happen? Um, thanks for coming on, Tyler.

Where can people find out more about you? Yeah, you can check us out at vocer.ai. Um you can see demos about the product and uh feel free to to grab a call on on our website uh with me.

I love uh talking to folks and solving business problems in all my roles before. I've been sort of a business data analyst and so where we always start with new customers is like what are the actual business challenges you're focusing on today like what are the initiatives that are going to move the needle for your business and then we bring in our expertise from talking to so many other folks who have used AI um to help you decide like where you should start and get that off the ground quickly with Boker. So I was Tyler