Category
Team
Perspectives
© 2026 Category Ventures
SF           
← All perspectives

How AI Agents Interpret Law and Regulation, with Ralf Schonherr, CEO and co-founder of Myriad AI

Çağla Kaymaz·Nov 2026·45 min

Ralf Schonherr is the co-founder and CEO of Myriad AI, where they’re building compliance agents that monitor and interpret regulation accurately at global scale. Before starting Myriad, Ralf got his PhD in self-learning systems and then joined a small team at Google X trying to teach software to write software, years before ChatGPT.

Listen in as we discuss the following and more:

  • What a PhD teaches you about finding product market fit
  • The Google X "monkey and the pedestal" mantra, and why you should de-risk the hardest part of your product first
  • Why code review is the new bottleneck now that agents write the code
  • Why hardware is 10x harder than software, from someone who built both
  • His take on 9-9-6 culture
  • AI-assisted interviews

Disclosure: Myriad is a Category Ventures portfolio company.

Also listen onSpotifyApple Podcasts

Introduction

Ralf Schonherr

If the problem that you're presented with is you want to have a monkey sit on a pedestal and cite literature — Shakespeare, for example — they always say, Start with the monkey.

And what this actually means is try to teach the monkey how to cite literature, Shakespeare, because if you can't do that, there's no point of building the pedestal. I think what a lot of companies tend to get wrong is they break the problem down, like bottom-up, top-down, you're going to name it what you like, but they build a pedestal and then they never figure out how the monkey can actually learn the hard part here. So I think starting with the hard part of the problem is something that we do at Myriad AI every single day. And it's part of the mantra we're trying to live as well where there's things we can do, we know we can do, but in order for this new direction to become a product, there's just certain challenges and if we will never be able to overcome them, we're at risk of wasting resources and engineering time in a direction that'll never materialize.

So that's actually a very powerful mantra. And I think that is true even for the early stages, right? If you have this idea and this crazy thought of this new company you want to build, I encourage everyone with this concept to start with the hardest part of your problem and almost set yourself up to a way where you come into work every day and you try to prove to yourself or to the world this can be done.

Cagla Kaymaz

Welcome to the Main Branch podcast. This is your host, Cagla Kaymaz. Main Branch is a show where I talk to founders and early adopters about building, deploying, and scaling world-class AI startups.

My guest today is Ralf Schonherr. Ralf is the cofounder and CEO of Myriad AI. At Myriad, they're building compliance agents for heavily regulated industries like insurance and banking. I've been super lucky to work with Ralf since last year when we led Myriad's seed round, and I'm super excited to have him here. Ralf, thanks for joining me.

Ralf Schonherr

Thanks for having me today. Excited to be here.

From Robotics to AI Compliance

Cagla Kaymaz

To get us started, why don't we talk about your background? So you did your PhD in AI and robotics in Germany, then you moved to the US and you worked at a bunch of different places, including Google X, and then you started two companies. Let's talk about that. And also I was just thinking, how did you end up in compliance? I probably asked you this, but your background doesn't completely scream compliance to me.

Ralf Schonherr

No, no, it doesn't point that way. So to look back, it's actually interesting for me. I felt a lot of my career path has been determined by the fact that I don't follow external advice. So I had a lot of people to tell me to go there, do this. And I think I've been doing much better usually following my own interest. So the way my story really unfolds is I started originally both business and engineering because I couldn't decide between both of those disciplines. And then as I was doing it, I really fell in love with math and statistics. And so I did double down, did my master's in systems engineering at Georgia Tech, actually. It's where I met my current cofounder 15 years ago. After all the math, I was kind of asking myself the question, Hey, what's the coolest part of the world you can now conquer with this?

And the cool part of math — and I'd argue it's all just math at the end of the day — it turned out to be initially industrial automation and robotics. So I started my career path there. The interesting piece is I worked at BMW for several years, was definitely at the right time and the right place, I believe more than once. So fairly early in my career, I grew with this large responsibility. I was, in the end, globally responsible for anything innovative automation at BMW. And that meant that all these very awesome, expensive, and super innovative systems would be my problem if they wouldn't work. And the general nature of those systems was that they would work really well if they were only doing the same thing over and over again and the world was static. But the reality is the world isn't really static, right? The world's continuously changing.

And I was very frustrated about the fact that there was no way to properly teach these systems to react to an unforeseen circumstance. So that got me really interested in the concepts of self-learning systems. That's about a decade ago. And I started reading about it. I got ultimately very obsessed with it. So I decided to go back to school in parallel to my full-time employment and I did my PhD in self-learning systems. That was kind of the opposite of what everyone recommended me doing, obviously. Because of this background, I was then discovered by a very small, very top-secret project back then within Google X that had the idea — eight people when I joined — to say, Hey, why don't we try to figure out how we can teach software how to write software? That sounded extremely appealing. So that's about eight, nine years ago.

So I got into this whole idea of ultimately GenAI fairly early on. We worked closely with DeepMind, saw some of the stuff I think before ultimately everyone else, but it was before the breakthroughs. So before the breakthroughs, and it was hard for me to see where this is going. So after a couple of years, I decided to leave, build my first company, and then shortly after, my second company. But yeah, I think in a nutshell, the reason how this ties ultimately to compliance is — and that's actually not on my resume anywhere — is with this sort of being unfolding for me, I still grew up in a family of lawyers. So it's almost funny to a degree, I'm like the black sheep in my family. I'm the only one among my siblings who didn't study law. So I think my sister, my three brothers, my dad, everyone's an attorney.

I think we even have a judge now in the family. So I grew up fairly adjacent to the problem. And I was always fascinated by the fact — given my modern world of technology — how antiquated those workflows are, how manual the work is, how expensive the labor is. And that was ultimately my way to having an understanding for the technology, but seeing the opportunity in the space. And that was the conviction that I needed to start a company together with my cofounder.

Cagla Kaymaz

That's really interesting. Actually, my husband's family is full of doctors and he didn't follow that path. After doing a couple other things, now he works at a health care AI startup.

Ralf Schonherr

(laugh) Yeah. I always say you tend to come full circle. It took me half my life, but ultimately I think now I'm here and I'm actually, I'm very excited to combine those two disciplines into a new kind of company today.

Why a PhD Is Like Building a Startup

Cagla Kaymaz

So from your PhD, I mean, you had a couple other experiences, but I'm curious, because I've worked with a lot of technical founders, I was just thinking through our portfolio, this wasn't intentional, but I think the majority of our companies have at least one cofounder with a PhD. What was that experience like? Were there anything from your experience in your PhD that carried over to starting a company, and what were some things you had to learn on the spot?

Ralf Schonherr

Yeah, that's actually a really fascinating question. So I know a lot of people who say this thing to me where it's like, Hey, I wish ... I want to start a company, I just don't have the right idea. And what I usually tell them is I refer to my PhD, because look, when you decide to do a PhD, you have a rough direction. You're fascinated, you think you see an opportunity, and you're very passionate about a particular domain space. So that's what happened for me. And then you sign yourself up for it and then you're going to put a title on the thesis of that's the work you're going to do. But the reality is you don't really find what the market or the scientific community needs until you actually start walking in that direction or running in that direction. So for me, I set myself up for like, I want to do this work ,and as I was investigating and digging, I ultimately found what was needed.

And I think to some extent that is very comparable to how building a startup feels like. Because you know the direction, you see generally the opportunity, but how to solve the problem, how the product actually needs to look like, how do you need to kind of position it, market it, price it, these are all things you will almost never get right first time. So the only way to ultimately discover this is you just go in that direction. You run as fast as you can, hopefully in the right direction, but you continuously reassess where do you actually need to position this. And to some extent, there's a lot of parallels between finding the topic to add to the scientific community in my PhD to finding product market fit with a company. And the thing that I tell everyone is you won't be sitting in a room by yourself and there will be this light bulb like, oh, this is the company I need to build and that's what it's going to be.

But you need to have, obviously, conviction, but then that's the thing that carries you in the right direction. And then from there, you'll find your path, but you almost never see it all the way through from the very beginning.

Cagla Kaymaz

I mean, I'm assuming when you were making this switch, a lot of people gave you advice on like, do this, don't do this, I've been in your shoes, or I've seen other companies do this, investors and founders and so forth. I was actually recently at HBS, Harvard Business School, and I heard a startup adviser tell students that they don't need technical cofounders anymore. They can just vibe code their way and why would you give equity to someone else to come and code for you, which is just a bizarre way to think about it. So I don't think every advice is to be taken as seriously. Have you ever been in a situation like when you were getting started, someone gave you advice and it ended up doing more harm than good?

Ralf Schonherr

I've gotten a lot of advice over the years. And I think part of this is people giving you unsolicited advice — and people tend to do that a lot — but then also I'm very actively seeking advice. So I have people, a community that I surround myself with that I continuously use to say, Hey, I have this idea, what do you think? It ultimately comes down to you making your own decisions, though. And I think the one thing that I felt specifically with the unsolicited advice, when I started my career path early on, there was just like, Hey, this is what you're going to do. You're going to go here. I had a mentor and very brilliant, smart person, but he was like, Don't go into this direction. Don't go and work on this innovative stuff. That's not the right career path. Don't pursue a PhD while working full time. But ultimately, looking back, those are probably some of the things I'm most proud of in my life and the things that set me up on the path that I'm currently on.

So I think for me, I haven't really taken any bad advice because I think my general nature is I just prefer to do what I like and I really follow my passion. And I think what is certainly true for me is if you find something you're actually truly passionate about, it's very easy to pour your heart and soul into this. And you also tend to be generally better at doing this because you care, right? So if you find what you actually are passionate about, I wouldn't say neglect everyone's advice, right? Obviously you want to make sure you hear people out, but ultimately, I think, I believe at least for me, the more successful path is following down your story, following down what are you passionate about and the things that you care. And ultimately that might be sometimes 100% opposed to what people tell you and sometimes it's a shade of gray of like what other people suggest you're doing, ultimately.

Cagla Kaymaz

And did I hear you correctly? You were doing your PhD while you were working full time?

Ralf Schonherr

Yeah. Yeah. And you can imagine there were a lot of people who told me not to do that, but ultimately the combination of the ability to be in the real-world environment. So I think what made this very successful for me was two things. I was in this role at BMW where I could see and really experience the real-world problems. And on the other hand, I still wanted to pursue the problem because I got so obsessed with the idea and the possibilities of it, and combining these two things, not just working on another toy problem. Because sometimes when you read some of these papers from the most reputable research institutions, they're still playing Legos. And I'm like, That's not the real world. In order to really make its way through, you have to think about this from a real-world use case.

And that made this very interesting, appealing, and I think ultimately most successful for me was I was able to combine it. But you can imagine, there were a lot of people — including some parts of my family — telling me absolutely not to pursue that. And I'm really glad I did at the end.

The Monkey and the Pedestal Framework

Cagla Kaymaz

Yeah. Yeah. I'm glad you followed your heart. I mean, sounds like you generally follow your own heart and what you think is right. Have you been in environments where there was advice or culturally you heard something, maybe at Google X, you're like, Actually, I hadn't thought about it in that context. Now that has changed how I approach life.

Ralf Schonherr

Yeah, look, I mean, I'm not saying I know everything. It's actually the opposite. I know that I know very little. And I think a lot of people who pursue their PhD will ultimately acknowledge that I know very, very little. It's more I go down the rabbit hole of a particular problem. But yeah, there's certainly a lot of mantras that I came across that I actually thought are extremely powerful and I like them to make part of my day-to-day, but also for the company that I'm building. And one of the things ... I think Google X is a great example. So there's a particular mantra there that talks about this example of the monkey and the pedestal. So the idea is if the problem that you're presented with is you want to have a monkey sit on a pedestal and cite literature — Shakespeare, for example — they always say, Start with the monkey.

And what this actually means is try to teach the monkey how to cite literature, Shakespeare, because if you can't do that, there's no point of building the pedestal. I think what a lot of companies tend to get wrong is they break the problem down, like bottom-up, top-down, you're going to name it what you like, but they build a pedestal and then they never figure out how the monkey can actually learn the hard part here. So I think starting with the hard part of the problem is something that we do at Myriad AI every single day. And it's part of the mantra we're trying to live as well where we're asking ourselves, there's things we can do, we know we can do, but in order for this new direction to become a product, there's just certain challenges and if we will never be able to overcome them, we're at risk of wasting resources and engineering time in a direction that'll never materialize.

So that's actually a very powerful mantra. And I think that is true even for the early stages, right? If you have this idea and this crazy thought of this new company you want to build, I encourage everyone with this concept to start with the hardest part of your problem and almost set yourself up to a way where you come into work every day and you try to prove to yourself or to the world this can be done because that is kind of the most ambitious way to get to a point that you really push the limits of it. And if you actually figure out how to do it, building the pedestal, that's the easy part, right? That's something you can do fairly quickly.

Cagla Kaymaz

I guess in Myriad's context, what are some examples of what would be the pedestal versus the monkey?

Ralf Schonherr

So I think in our instance here, what we've seen is that there are certain aspects of the product that are heavily on integrating and workflows — UX, UI — and that's what I consider the pedestal. There's things we know how to do. There's like, how do you assign a ticket and enable collaboration? The one thing when we started a company — and remember, this is like mid 2023, late 2023. So there was a lot of uncertainty around capabilities of the models. And the one thing we had to discover is, can we actually interpret law and regulation in a way that is comparable or better in its accuracy than what the human mind can do? That is the big question mark. And so we invested actually quite heavily in the ability to do two things. So when we're taking a law and a regulation, we're completely rewriting it into a new concept that we invented that allows the law and the regulation to become a lot more accessible to the interpretation of the latest and greatest models.

And the great thing is as they get smarter, we're getting even better. And then the second piece is you need to have a lot of context, right? So in order to interpret oftentimes a paragraph in a regulation or a law, it's very gray, right? It says, you can't indirectly do X, Y, Z. And then what does "indirectly" even mean? In order to understand what this means, you need to look at real-world enforcement, others' examples, cross-references. So building a context tree, but then also making sure you only feed in the relevant piece of these massive context trees, that was a big part of what we needed to discover and uncover. And honestly, I remember talking to my wife in the early days and she was like, Do you think you can do it? And I'm like, I don't know. I really don't know. But we invested heavily into trying to find out if we can do it.

And as we were able to show evidence that this can be done, that was the moment in time where we're like, Okay, I think we've got a company, we're going to scale this now, and we're going to start us running as fast as we can.

Balancing Speed and Accuracy in AI

Cagla Kaymaz

You mentioned in your industry, right, accuracy, security, data freshness, reliability, these things are super, super important. Unlike some other AI startups, you can't just vibe code and say, Hey, we'll fix it as we go, because these industries are very risk averse. But then at the same time, you do need to balance that with shipping fast, iterating, getting customers. How do you think about balancing those two?

Ralf Schonherr

That's actually a really interesting one. So in our case, I think there's what I call table stakes, which is this metric of precision versus recall, where you need to be extremely accurate. We cannot have things slip through and say, Hey, this is totally fine. But it's not. So that's part of how our tech stack is set up that we're erring on the side of being extremely cautious, and that gains trust. The other thing though that we have to ask ourselves continuously every month is how do we benefit from the speed of technological progress? And part of this is how do we create software? And there's an interesting, I think, moment in time right now — and really since the last release of the most recent generation of the models — coding capabilities have significantly increased. And we're now very actively challenging ourselves in our ability to do code reviews.

So now as you can generate lines of code so much faster, someone needs to read this. Someone needs to make sure this is actually not spaghetti, right? It's actually making sense. It's computationally efficient. And that creates a real human bottleneck. Don't get me wrong, I think if you want to build an enterprise product, you will not be able to vibe code it end to end because it ultimately requires a different sort of access. But what we're definitely experimenting with is modular units that we can kind of swap in and out. These are heavily created with GenAI coding capabilities. But now as they live, how do we ensure what's inside of it is actually what we want it to be, and on the other hand, how do we do it in a way that it doesn't significantly slow us down? How can we actually be 50, 100 times faster than what a company was able to develop three, four years ago?

So that's a big challenge. That's actually totally top of mind right now. And I think the entire world is trying to figure out how to do this in the most efficient way right now.

Cagla Kaymaz

One of the things I've been super impressed with, you and your cofounder, Jan, is every time we think there's just some new update in the product that I didn't think was possible in that 10 days we hadn't talked or something like that. That's been really impressive. And I don't think you can be an AI-native company these days without actually being a power user internally. Sounds like code review is one of them. Are there other instances where you guys are utilizing AI significantly internally?

Ralf Schonherr

Yeah, we use it across the stack everywhere. Our ticketing system is like ... We're having a conversation in Slack and then we're saying, Hey, make a ticket out of this thing. That saves a lot of time. That's so trivial, but that's a huge time saver, right? Just the entire conversation, not enough someone needs to click into this particular tool and then create it and assign it. We're using it heavy on the code generation. We're experimenting with code review, but we also use it, I think, in a lot of other ways when we think about experimentation, go-to-market, these kind of things, there's so much that can be done today. I play around with AI agents. I actually ... So part of what I like to do is I like to inform myself around a particular set of news pages every single day. And what I ended up doing is I created a couple of AI agents that scrape these for me and summarize them and just break it down to two to three articles that I care about and summarize them for me and they send me an email.

And again, this saves me half an hour to some degree. So it's personal, but also I think as a company, we're trying to live it more and more every single day. And it's actually a challenge because what was possible six months ago is very different than what's possible today. So you have to continuously make sure you inform yourself, you stay up to date, but then you're still trying to ship real products. So that almost is a conflict of how do you allocate your time into just informing yourself? But yeah, no, I agree 100%. The world has changed and it's continuously changing, and if you don't benefit from it, you are at a disadvantage.

Cagla Kaymaz

I don't know what your morning commute is like, but you can also have a voice agent read those updates to you to even further —

Ralf Schonherr

I'm going down Market Street on a very sketchy eScooter and I would not recommend doing anything but paying attention to the traffic there.

Why Hardware Startups Are 10x Harder

Cagla Kaymaz

Yeah, I probably would not have AirPods on on that commute so I take that back. I guess going back to the beginnings, you were one of the earlier self-learning, you studied self-learning, you worked in robotics, and then you actually started a company at the intersection of robotics and AI, and it was a more vertically integrated hardware and software play. And then now you're only doing software. I mean, as you know, there's a lot of investment going to robotics right now. What's your perspective on it? What was it like to work in a vertically integrated startup versus software only, and why did you make the switch?

Ralf Schonherr

So I applaud everyone who tries to build a company that has a hardware aspect because that field's 10 times harder than pure software. And I think the predominant reason for this is because you actually develop at different life cycles. So when you imagine how quickly you can do software versioning, that can be weeks, right? Maybe two weeks, as simple as this. When you go then into the hardware world, you quickly realize, Hey, so if we're going to get a new generation of a hardware product out within nine months, we're really fast. Because sometimes you go and you order some of these pieces that you need for your hardware components and they have six months lead time to even get a first part. And you're not sure that's actually the right part and you want to do some testing on the hardware and do some iterations.

So it is incredibly hard to get these two life cycles synced because ultimately hardware-enabled software is a very exciting place to be in because there's things to be said about defensibility, there's things said about things you can actually do that you cannot do purely with software, but it makes essentially your life very difficult in the ability to be fast, iterate. And on the other hand, it's also something that I think people tend to underestimate how expensive that is, just like pure development costs. And then now you want to scale, so your cost proportion scales to your deployments. So that's another one where you need to make sure you have enough capital reserved to even just like, let's say, hold 100 units ready to go to the market.That's a significant amount depending on what tech you're actually developing. That's a significant amount of money in your bank account that is now sitting in hardware that might not even be the right thing you want to do because you're already waiting for the next generation.

So I think to that degree ... I've been there, I've done it. It's really freaking hard. I applaud everyone who does it because I think the world needs it, but to be honest with you, it was almost a conscious decision after being related to hardware for so long, I wanted to make myself ... This is hard enough. So not having to deal also with this hardware life cycle felt like a way to keep a little bit more sanity and not make this unnecessarily harder than it already is.

Cagla Kaymaz

I'm sure the founding experience there, a lot of those skills transfer, but it sounds like a very conscious decision to not have a hardware component at Myriad.

Ralf Schonherr

Yes, no, no. First of all, we don't need one. And I actually really like ... So the way I approached this, and I think there's a difference between my first and my second startup. The first one, I think I might've been guilty a little bit of like I had a hammer and I was looking for a nail. With the second one, very organically, it was like, Hey, I see this opportunity. I see this problem that everyone tends to have. I know how the tech needs to look like, at least at a conceptual level, for it to work. So that [drew] me into this and luckily no hardware required. But on the other hand, yes, I also would've been very hesitant after so many years dealing with the pain and eating glass of the hardware development cycles to do this to myself one more time, to be completely transparent with you.

Coding a Startup Prototype on Honeymoon

Cagla Kaymaz

So actually before today, I was talking to your cofounder, Jan. I was like, Jan, is there anything that you think I should ask Ralf about the founding journey that I might not already know about? And he said to ask you about your honeymoon. What happened there? What's the story?

Ralf Schonherr

Yeah, I think it says probably a lot about me. Luckily, I think I married the right woman. So the part here is when I saw the opportunity that I just talked about, I started playing around with the first prototype, so what ultimately turned into Myriad AI. And I had the time to actually do that during my honeymoon in Italy. So we were there for a couple of weeks traveling. And yeah, when everybody was asleep in the night, I would sneak out of bed, pull out my notebook, and actually start creating the first prototype of what then turned into Myriad AI and ultimately plugging those pieces together. And I'm very glad I did because ultimately here we are. But also to be honest, I have a lot of appreciation for my family who's been tolerating me. I think the other way someone described me as a very driven kind of founder type.

And I'm still not sure if that was a compliment or not, to be honest with you, but I think that's part of who I am. But also, yeah, I think you need to ... And maybe that's another thing that needs to be said. In order for this to be successful, you need to have a support system that allows for that. And part of this is your wife not threatening you with a divorce if you actually start writing the first prototype of what turns into your next company during your honeymoon.

Cagla Kaymaz

Yeah. At the time, it sounds like she didn't do anything, she didn't say anything, but in hindsight, do you guys talk about this, and what's her reaction these days?

Ralf Schonherr

She's still talking about that for sure. No, no, no. She was very much aware of what I was doing. And I think for a really long time, she's like, Look, this is how much I'm tolerating it. And I'm like, I'm very grateful for it. But yeah, no, I mean, we're both glad I took that path because ultimately the way the company is trending, what's ahead of us, I think this is all proven to be a really good decision. Also, to be honest with you, that was a different time. So we're talking about mid 2023, right? So there was a lot of stuff that was just not even as accessible. So even just connecting and just plugging these pieces together, that was extremely exciting to me, but on the other hand, the world didn't even ... There was a time when people hadn't heard about ChatGPT necessarily yet.

So that was very interesting early days, but I'm glad I had the opportunity and the ability to do that. Would I want to do it during my honeymoon again? We'll see. Probably yes, but don't tell my wife.

Cagla Kaymaz

You have a history of doing things early before others catch up, but at the time, was Jan already with you or this was before he joined you?

Ralf Schonherr

Funny enough, so he is my CTO, cofounder, I tried to convince him to actually help me with this as I was playing around with it during my honeymoon. I was like, Oh, I'm trying to figure out ... And he was very busy at this moment in time in his life in work and everything so I actually couldn't get him to help me. And I'm still kind of rubbing that in every once in a while, but I was able to figure it out by myself, but it was kind of like he could obviously tell from what I was experimenting with the direction that I wanted to go. And we've been friends since 15 years. So he actually was at my wedding, right? So he was there and obviously we had a chance to connect and talk about it as well and I might've hinted to the fact what I'm planning to do.

So I think that might've been the first time that I socialized the idea of like, Hey, what are you up to? Maybe there's a chance for us to work on something together. But yeah, it was definitely the very, very early days before we even put the company on paper.

Cagla Kaymaz

Was this when Jan was at Airbnb or Nvidia? What was he doing in his life?

Ralf Schonherr

He was still at Airbnb. So he was essentially the tech lead for the service framework team, which is powering the entire back end to Airbnb. But yeah, honestly, talking about him for a minute here, I am so impressed with his career path, somewhat similarly to finding he has such a good sense for joining companies at the right moment in time. So when we departed Georgia Tech and I took my path, he went to work at Nvidia. So this is like 15-years-ago Nvidia, so no one ever heard about it. It was not an attractive company to join, but he had the right feeling for it. And then from then went to Dropbox, IPOed with them, spent some time at Google, and then also joined Airbnb before it was such an obvious success. So I think to that degree, I always say he has a really good sense for joining companies who are then very successful.

And I hope that's true for Myriad AI as well. But I think that's probably one of his superpowers of knowing where to go at the right moment in time.

Pros and Cons of a Distributed Engineering Team

Cagla Kaymaz

I'm so glad he ended up joining. He's phenomenal, big fan of Jan. And eventually after he joined, now you're in SF, but you do have a large R&D team in Prague where Jan is from. I've been super impressed with how you guys have been balancing multiple geos, time differences, having a distributed team. Why did you make that decision to build part of the team there? And then what are some of the ways you guys make it work? Maybe some tactical advice for anyone that's looking for European R&D plus SF Bay Area core team.

Ralf Schonherr

So objectively on paper, the reason why that can be a pretty good decision is, so you gain a lot from it in terms of your access to the talent pool. Just having to recruit in one particular city such as San Francisco, you now have this other hub where you can actually access talent. And interestingly, so if we're trying to hire people here, we're kind of competing with the Anthropics and OpenAIs of the world and they're paying insane salaries. So being a startup and saying, Hey, even if we're giving you this much stock, it almost makes no sense from a business case perspective for the actual individual to come join you. Then you go to a place like Prague that has an incredible technical university, so you have top talent there. There's no Anthropic or OpenAI office, at least not to the best of my understanding.

So now, as Myriad AI, we're one of the hottest startups in the city. People actually want to come join us and we're actually, we're able to pay fairly competitive salaries, and that's still a fraction of what the cost of an individual engineer would be in San Francisco. So what it allows us to do is we're getting really the top 1 or 2% of the talent pool at a lower cost. And then I think the additional really big upside is we're operating in every time zone. A big part of our existing customer base and our pipeline also is European based. So being in the same time zone and able to react, that makes a huge difference. It's not like an outsourced customer support. It's literally Myriad AI engineers in your time zone responding to anything you need and you want. So that's on paper everything that makes us a very good and appealing scenario.

The thing though that I caution everyone who thinks about it is that you need to be able to do that, and it comes with sacrifice. So the first thing you need to be able to ensure is you have someone on the ground that you truly trust. So in our case, me and my cofounder, we're talking every single day and we're spending hours with each other, but a lot of that is actually video chat because I'm in San Francisco the majority part of the time and he spends a lot of time in Prague, but he needs to be on the ground. So if people were to say, Hey, I'm going to outsource this to a team in a country where I'm not at, where I don't understand the culture and I'm not even there, that tends not to work. So it needs to be really your core DNA, it needs to be your office, and you need to have someone there that you trust entirely.

I think that's almost a non-negotiable requirement for this to be a very successful solution. On the other hand, yes, time zones have challenges. You need to be very smart about calendar management. So I get up super early. A lot of my days start at, like, 6:30. I'm okay with this, to be honest, I've gotten used to it over the years. But then I think there's more subtle things. How do you ensure you feel as one company? How do you ensure communication happens? So one of our core principles within the company is we're over-communicating, but over-communication doesn't mean we're just communicating a lot. It's the speed of communication. So we're very concise in our communications, but we're very fast, very rapid. We're using a lot of these toolings that are existing today.

And yeah, that's something you have to worry about every single day because otherwise you create two separate companies in silos or you actually end up being slower than what you want to be. But on the other hand, the good thing is, as my team there at Google grew, we had that scenario relatively quickly. We actually acquired a company in Germany. So same thing, San Francisco, German time zones. My other startup had the exact same setup. We had an office in Munich, Germany, relatively early on. So over the years, I think I've gotten really good at learning what works and what doesn't work, and that makes us more efficient for Myriad AI.

Cagla Kaymaz

What has been the most difficult part?

Ralf Schonherr

I'd say the most difficult part is I think really actually ensuring you stay in sync. So if someone needs to wait nine hours for you to come out of bed ... And that's kind of the worst-case scenario. So Prague gets up when I go to bed. There's a little bit of overlap, but not a whole lot. How do you ensure that doesn't cause delay? How do you ensure you empower people in their particular teams to make good decisions on their own? If they need to wait for someone in San Francisco to come up or think on something. And I think that's a really hard part to get right. I think we're doing this really well now, but yeah, that is, I think, probably the biggest challenge because that can completely backfire, because what can happen if you do it that way and people are not empowered and communication isn't properly shared, then people might end up waiting nine hours until the end of their workday, Friday, to get the info to now work on something and you actually effectively lost a day.

So I think it's a slippery slope. So if you do it well, it's incredible and powerful, but it's also, it comes with the challenges and it can kind of cause the opposite of what you want it to ultimately bring.

Cagla Kaymaz

Now you guys have a team in Prague, you used to work in Germany. I love Silicon Valley, big fan of SF. I think this is where most companies should be built. There's exceptions, but big fan of the culture here. But in your experience, have you noticed anything different between Europe and Silicon Valley, and what's your reaction to some of the more exaggerated parts of the culture, like 9-9-6 and other things?

Ralf Schonherr

Yeah, that's actually interesting. I think the Czech Republic, similar to Germany, tends to be a bit more conservative. And we're trying to actually make that a selling point. So we're this warm, fancy, hip, Silicon Valley-based startup in Prague. So there's a lot of things where you have to [break] up the mindset and these things like ask questions during all hands, people tend to be a little bit less comfortable. But on the other hand, yeah, I think the Silicon Valley/San Francisco Bay Area is almost doing the extreme opposite of it. So you run into these mantras. And I think that's this thing that sometimes might drop me off a little bit the wrong way when you talk these cliche conversations of like people saying, Hey, we're doing 9-9-6 and all of this. I think it's ultimately ... The way I want to compare this is when the idea of radical candor came out and people were like, Hey, excuse me, I'm radically candid, then they were just using that as an excuse to ultimately be an asshole, to be honest.

So I think to that degree, some of the stuff you see here is the same thing. Look, I'm not saying ... You need to be really fast and work extremely hard to build a company, right? But I think some of these mantras that people then really emphasize are actually misused. It's the same thing with radical candor and 9-9-6 now becomes like a mechanism to squeeze the last juice out of your employees. And the fact that you spend a lot of time in the office doesn't mean you're doing the right things. And I think time in the office is for me the wrong metric, to be honest. We're spending a lot of time in the office, don't get me wrong. We're working hard and there's weekend work required in all of this, but I think it depends a little bit on how we live the mantra. I think the better way to think about this is your company and the people in your company are assets, and what you want to make sure is you want to treat them well, you want to keep them happy, you want to set them up for success.

For some people who are often more senior, that means you need to be able to combine both, like pick up your kid from day care and still be able to get your work done and all of this. So I think to that degree, I sometimes trip over some of these more extreme ways. And I think San Francisco tends to be also a bit polarizing when it comes down to this tech culture. But yeah, that's certainly not our way.

Rethinking Startup Hiring and Interviews

Cagla Kaymaz

From a team perspective, have you recently been in a situation where someone wasn't a fit and how did you handle ... How long did it take you to figure that out? How did you handle it? How does that now weave into your hiring practices to get that early?

Ralf Schonherr

Yeah, look, it is actually an important one. And everyone says hire and fire fast, and it's like another one of these Silicon Valley mantras. But look, I don't think it's necessarily wrong, but it comes down to, first of all, you need to make sure you hire the right candidates. So our interview process is actually pretty extensive. Actually, candidates acknowledge that. So you have to go through five, six, seven interviews, and they're hardcore. Actually, it did a really good thing for us because we're now known to be such a hard gate to pass, people actually want to come and work here. There's like, Hey, there's this challenge. I want to get into this company that it's really hard to get into. We've also had ... Another founder actually came to me, he said the headhunter approached him with a candidate and he said, Hey, this is a really good candidate. He made it to the final stages with Myriad AI.

So it becomes now a quality step for us. So that's one part to actually ensure you don't have to do that a whole lot. Really ensure you get enough data points in the assessment. That being said though, yes, it does happen. Sometimes you bring in someone to do a particular job and you realize it's not working. And then the way we're approaching is we're actually paying very close attention to all of this every single day with the structure that we have. I talk to everyone in my company almost every week or every other week. And once we're surfacing this, what we're doing is essentially we're understanding, hey, why is it not working? Is it like what's in the way? Is there a roadblock we can clear? We're trying to empower the people to do the job they can't do, but if we feel like they just can't get their role, so to be honest, fairly quick on making a decision and say, Look, we'll be fair. We'll try to set you up and have a soft lending, but this is obviously not working in terms of what we need you to do. And yeah, we're not afraid of making a tough decision there as well.

Cagla Kaymaz

Actually, this reminds me, I was talking to one of my other founders literally earlier this week and he was saying there's a lot more fraudulent or at least gray-area behavior happening on video calls. For example, they were interviewing someone for a technical role, like an engineering role, and after asking a certain question, there was a pause where you could kind of tell he was using some sort of an assist on the side to help answer the question. And obviously there's some of this happening in the resumes and job applications. Are you guys seeing any maybe AI-generated or AI-aided behavior in these interview processes?

Ralf Schonherr

Oh yeah, 100%. Look, I've been in interviews myself where I ask someone a technical coding question and then they kind of fluff around for like 30 seconds and then they give you the perfect answer, and you can tell what's happening. So that's part of the reason why we're doing onsite interviews. So we're doing the initial technical phone screen, maybe another one if we need more data points, and then people have to come and meet us in person. That's how we can ensure that this is not happening. But yeah, actually it's an interesting one. So I think if you plan to build a purely remote company — luckily we're not — but that's probably a huge challenge there, to be honest. How do you ensure that people are actually not giving you the results of an assistant that's running on the side? But yeah, for us definitely something we witnessed and it's part of the reason why you need to spend a hardcore interview day with us in order to make it past the quality gate there.

Cagla Kaymaz

Have you yet had anyone that faked their way through the virtual interviews and then onsite they completely bombed?

Ralf Schonherr

Yeah, yeah. We've had that happening for sure. But you can usually tell. There was also like, there were cases where the candidate acknowledged that and halfway through the interview day they were like, Okay, I guess I'll just go home now. We literally had a candidate just leave during lunch without saying anything. I think that candidate particularly acknowledged that this is not working. And yeah, I think to that degree, that does happen. Yeah, 100%. It is kind of a new challenge with everything that's going on in the world.

Cagla Kaymaz

Yeah. Unfortunately, I don't think it's going to get better, right? So these onsite interviews seem to be the best way. And obviously even during onsite, you probably want them to be AI-first ,using all the tools in their disposal and stuff like that, but they also need to have the fundamentals. They can't just give you the right answers without knowing how to do the job. I want to end this with some quick-fire questions. I have three of them for you. Are you ready?

Ralf Schonherr

Ready.

Cagla Kaymaz

On a Sunday afternoon, if you're not working, what are you doing?

Ralf Schonherr

So I think my answer used to be, I love surfing at Ocean Beach or riding some gnarly downhill mountain bike trails in Pacifica, which by the way, is great terrain. If anyone's interested, I can show you. They're not on the map. No one talks about them officially, but the community knows them. Now with two kids, I spend a lot of time on playgrounds, to be honest with you. So it's either-or, to be honest. And there's some really cool playgrounds that, believe it or not, San Francisco and the Golden Gate Park has the first playground in the United States.

Cagla Kaymaz

Oh really?

Ralf Schonherr

The one that is there. Yeah, I walked by the other day. So there's some good stuff there. Yeah. But I spend a lot less time doing these things and not the other, but it's equally as good, honestly. It's very fun.

Cagla Kaymaz

Yeah. Until they're old enough and then you can take them to Ocean Beach to surf.

Ralf Schonherr

That's the plan. That's definitely the game plan. Yeah.

Cagla Kaymaz

Yeah. And the second question, best purchase you made in the last year.

Ralf Schonherr

The one thing that I keep talking about, and my wife almost makes fun of it, is I purchased one of these really fancy water filters. They have this osmosis technology. I have really good-tasting water right now at home, so I actually drink more water. I tend to be dehydrated. And the other thing, truly objectively, when you live in a place like here, you always worry what's in the water actually, can you give the water to a newborn and so on. I mean, I know that you don't give water to a newborn, but like later on you start when you introduce water. So I think that's one of these purchases that I've been constantly talking about. I'm like, I'm really glad I bought this. This is great. This is fantastic. But yeah, it's as trivial as good filtered water.

Cagla Kaymaz

Send me the link! I'm intrigued.

Ralf Schonherr

I can send it to you. 100%. Yeah.

Cagla Kaymaz

And last question, I don't know if you listen to podcasts or read, but what's something you're reading or a podcast you're listening to these days?

Ralf Schonherr

Yeah. I wish I had more time to read. I think given my work schedule, it essentially extends the entire day from the morning to the evening. I literally close my notebook and go to bed and then there's family time somewhere in the middle that I squeeze in. So I don't have a whole lot of time to read and I'm trying to actually ... I think part of what I want to do better is finding more time to read. With podcasts, it's equally as hard for me, to be honest with you, for two reasons. One's when my ears are somewhere where my eyes are not, I usually don't do really well. So I put noise-canceling headphones on and I need focused work. I can't just listen to something while working on something else. My brain's not routed that way. That being said though, I like the Acquired podcast.

There's a lot of books that I have in my shelf that are all about improvements. I don't really read a lot of stuff that is not just useful to me. I think there's these type of people who like to read stories. And don't get me wrong, I like science fiction, but I don't really find time for it. The books that I read are like "The Hard Thing About Hard Things," right? And these kind of things where it's just like, there's some philosophy or some advice, but yeah, honestly, I wish I had more time for it. The other thing that makes this challenging is with the remote time zones is we're often recording meetings that happen in an early time zone. And then I actually spend [time] running the recording better than the podcast to make sure I'm catching up with some of the things that happened in the different time zone.

Cagla Kaymaz

Yeah. "Hard Things" is a good one. I read that a couple of years ago. Highly recommend.

Ralf Schonherr

Yeah, for sure. There's a lot of good stuff in there. I think it's not a playbook. There is no playbook. And none of these things translate one-on-one, but I think there's a lot of truth in that that translates well to what you ultimately need to do as a founder.

Cagla Kaymaz

Awesome. Well, this has been a fascinating conversation. Thanks so much for joining me.

Ralf Schonherr

Yeah, thanks for having me. This was really fun and yeah, I'm glad I had a chance to be here today.

Chapters
← All perspectives