Ep # 110: Engineering Success: Transforming Industries with AI ft. Dirk-Jan van Veen (CEO and Co-founder, Query Vary)

Adil Saleh (0:00:03) - Hey, greetings, everybody. This is Adil from Hyperengage podcast alongside Taylor, the co host and a very unique guest. Most recently, we started getting interest a lot of AI powered platforms, more so for developer teams and programmers and engineers. All of these. And today we have Dirk, who's the co founder at Query Vary. That is a platform that helps to validate LLMs at scale, to better prompt and to have a quality output. Alongside that, he's also a co founder and CEO at Syncware. That is slightly for a different industry. Like he's helping warehouses increase their productivity by using AI. Remember, this guy has done recently, PhD in geophysics, is kind of a geophysicist, turned into computer scientists, and now a whole lot more entrepreneur, co founding two platforms. Thank you very much, Derek, for taking the time. Dirk-Jan van Veen (0:01:08) - Yes, one way to smoke correction. My master was in geophysics, my PhD was in computer science, AI. Adil Saleh (0:01:19) - Oh, cool. So he's more of a computer scientist, turned into an entrepreneur, and he's had his PhD in the computer science, and he has done his master's in geophysics, is that right? Dirk-Jan van Veen (0:01:34) - Yeah, that's right. That's correct. Taylor Kenerson (0:01:36) - Perfect. Adil Saleh (0:01:36) - So, thank you very much, Dirk, for taking the time today, and we are so thrilled to have you on. Wonderful. So, getting started, Dirk, of course, thinking of someone with such a background, and of course, you did your masters on purpose. You wanted to learn how this AI is evolving, and with someone having experience and learning science and physics and what is the real connection with artificial intelligence. What was your purpose while doing your masters on top? You wanted to do PhD in computer science. What was that one thing that you thought, okay, if I do this, this is something that will help me acquire this, and this is what I'm going to do while you're just co founding a platform. Such as. Dirk-Jan van Veen (0:02:31) - So, at the end of my master's, I was doing it with a company called Schlumberger in Cambridge, and all the people around me had PhDs, and I felt like they had a more structured way of thinking and I wanted to get to that level, so that's basically why I enrolled in a PhD program. I was also working on a topic called hydraulic fracturing, which is a bit of a controversial technique to get oil out of the ground. And I felt like I didn't want to have a career in destroying the planet. I wanted to move into something that was more innovative and that had a better future. Adil Saleh (0:03:24) - Interesting. And of course, while you just studying and doing your research on this PhD in computer science, you were also building synchro so how did it help during your journey at synchro? This academics part, and of course the research part. I know LLMs are quite simple on a high level, but when it comes to serving different use cases, especially helping warehouses, it might get complicated. So how someone like you with that academic background actually elevated that. Dirk-Jan van Veen (0:04:04) - So actually they're very separate. So I did my PhD, I completed it, and then I wanted to have more impact than with just writing papers. So I decided I wanted to start a company and put that research into practice. And so I joined a co founder matching program called Entrepreneur first in Singapore. And in the first three months, you basically just have to pair up and find a co founder and some early customers. And if you do that well, then after three months they'll decide whether or not to invest and we are very happy to receive the investment from them. Adil Saleh (0:04:52) - Interesting. Entrepreneur first? Taylor Kenerson (0:04:56) - Yeah, I think. Where is it located, Dirk? Out of. Dirk-Jan van Veen (0:05:02) - So for the first three years we were operating out of Singapore, but we recently made the move over to the United States, focusing on the US market. Taylor Kenerson (0:05:14) - Love that. Amazing. Interesting. I kind of want to take it into a little bit you touched on when you went to get your PhD. One of the reasons was because you saw others that had a PhD, but you mentioned something about structured thinking and structuring your thinking. I'm really curious on how the structured thinking and getting your PhD helped you in your two companies and structure them, and how you went about bringing them to market and bringing them to different audiences, especially in an evolving niche like LLMs that are growing and have not completely fully evolved yet. Dirk-Jan van Veen (0:05:51) - So the jump from PhD to becoming startup founder is very unusual, and there's only a few things that I would say that really help in that journey. As a PhD, you have to write this thesis, and there's a lot of writing involved, and I'm taking the commercial role in the company. So the writing skills help transfer somewhat to being able to communicate more effectively in marketing. And another key skill for PhDs is to see it through to the end. So some perseverance. This is also what I think you'll need as a startup founder. And then there's obviously the deep domain knowledge. So my PhD, I touched on all the, I want to call it like classic AI theory. Like these transformers only were invented in 2017, and they took a little while to ramp up. So I think the main key advantage from having done my PhD in AI is that I'm able to catch up very quickly with the latest developments and just have a fundamental understanding of how the technology works. And what might. Having an intuition on what would perform well and what might not perform well. Taylor Kenerson (0:07:36) - Yeah, and it also sounds like, too, when you're doing your PhD and you're completing that process, like a lot of other things in life, whether that's athletics or maybe you are a military veteran, there are these characteristics and these skills that you need and that correlate to being a founder and an entrepreneur. Just like you touched on that perseverance, that ability to structure your thinking in terms of how you're delivering your marketing, how you're building your product, even your business. So everything really does touch and connect the dots to each other. So you also mentioned that you went from the hydraulic fracturing and you were like, okay, I don't want to get into this industry. This is shit. It's going to ruin the earth. So what made you go into, like, I know you said computer science, it was really interesting. But taking into syncware and then what you did with vary, like, how did you get into the LLM side of this industry? Dirk-Jan van Veen (0:08:28) - So Syncware was the first part. We were an AI company, and you don't found a company by yourself. I did it with my co founder. And so you need to take the strengths from both founders. So my co founder, he's an expert in robotics. And our initial idea that we got when we went through the entrepreneur first accelerator was to connect, to make robots of different brands collaborate. And so first you need the robotics expertise to be able to interface with all these different robots. And then the second part is, we're going to use AI to optimize the fleet. And we were focusing on the warehousing industry because that industry saw the fastest adoption of robots. And we were a bit naive going into that industry as complete outsiders. We thought, goods move in, the goods get stored, and the goods move out. Surely there's a high overlap, or like, there are synchronicities between all these different processes of moving material in a warehouse, where it turns out that it's much more complex and much more heterogeneous and it's not something that can scale really well. So that prompted us to change direction two years later. Taylor Kenerson (0:10:02) - Dirk, I just have one question before you continue as being a co founder. That's not the expertise. Like the expert of that domain, how do you balance that relationship and that perhaps maybe imposter syndrome sometimes in conversations, and when you're sharing your ideas with the other founder, that is the expert in this domain, and how do you work on that synergy? Can you just answer that as you continue on in your answer. Dirk-Jan van Veen (0:10:31) - So my co founder was an expert in robotics, so he had that part down. Both of us were not experts in material handling. That's what they call it, like basically warehousing. And so we had to learn that, I would say it's quite a tough industry and we had trouble selling in it because there are so many stakeholders to manage when you want to sell, to basically sell robotics project, because the warehouse owners, they don't know anything about robotics, so they'll basically say, I want a turnkey solution. You just build everything for me, just take care of everything. That role is a system integrator, and that system integrator is going to tie us with all the robotics companies. They're going to tie us with the software vendor. There's usually a warehouse management software in that company. They're going to take care of the maintenance, they're going to take care of the whole tendering process. And so as a startup, it is very difficult to insert yourself into that process, especially as you're the middleware, because you need access to different levels of the robots from these different robot manufacturers. So it's a game of managing too many stakeholders that I think is hard to lift as a startup. Adil Saleh (0:12:08) - Interesting. So I just want to touch a little bit more on Curi way, like talking about LS applications and prompt engineering, which has at the same time, it has become easier for developers to quickly get into the solutions and all, but at the same time, it has become complicated for them to refine those engineerings and now it has become quite a category. Of course, it's not as big as the other categories, it's evolving. And none of these products that have achieved a kind of a letter that makes it a unicorn business, but it still is volume. So how do you see this as a category from a commercial standpoint? Like, how do you see Curi evolving to larger teams and slightly complex use cases when it comes to refining the outputs of the different LLMs in comparison, we already spoke to one platform that is helping to compare with all three or four LLMs and get the best cumulative output for their problems. So how much of this is tailored more for the engineering use cases, like software engineers, full stack developers and all? And how do you see it commercially evolving for you as a business? Like in terms of your customer segment, in terms of revenue, acquisition, all of that? Adil Saleh (0:13:42) -  I just wanted to touch deeply on the commercial side of this category, this business where you're playing like LLM applications, comp engineering, who's going to do the first bigger impact, like in terms of efficiency and all of that? Dirk-Jan van Veen (0:14:06) - Okay, so I think the biggest thing that companies and businesses are focusing on right now, which is even harder than the technology, is finding the right use case for AI in the business. So finding something that streamlines an existing workflow or augments what the company is doing. And the people that know about these are very often domain experts in the company, the people that are executing these tasks. And then it's up to them when they've defined this thing that where they want to apply large language models. Adil Saleh (0:15:16) -  I was just asking what is your insight on the industry of LLM applications helping prompt engineering, fine tuning the prompts for developers? Who do you think right now is going to be your, let's say, best competitor? Or maybe what is going to be your next move as a business? What initiatives are you taking in place in order to make it scalable to larger teams? This platform, I'm talking about cure. Dirk-Jan van Veen (0:16:00) - Well, the biggest problem that people face when they build large language model applications is the reliability. So it's relatively easy to create a prototype, but then a prototype that works maybe 30% of the time. It works for the demo, it works for a few cases, but if it actually has to go into production and perform a commercial task for the company, then you need to have at least 90% performance and then a human in the loop. And eventually you will want to get to more than 99% reliability. That step is very difficult. And what companies are figuring out now is that they can't just eyeball it. They need to have a process, so they need to monitor the traffic that's going through their prototype. They need to know what questions people will ask to this AI. And two, when you know that, then you have to measure it. So you want to be able to test. Like on all these previous questions that we got, the performance is this. So it gave a good answer in 30% of the time. It cost this much. On average, users had to wait 15 seconds in order to get an answer. And then when you have that, you can craft a different prompt or choose a different model and see and back test if that new change made a difference, a good one or a bad one. So you need to have a process, you need to have the monitoring, and you need to have the tools to display and compare and put that feedback back into the model. And this is where a lot of education still has to happen in the market, where people, one, they don't know that they need this process, so they'll first need to realize that, or two, that they don't know that there are platforms out there like queryvary that can solve these things for them. Taylor Kenerson (0:18:22) - So it's very much largely to a lot around education in the space and bringing this knowledge to people that are not the technical leaders or the gurus on the team. And really letting the common Joe and Jane understand products like query vary and other instances where they can use and leverage artificial intelligence in the best way possible. And not just for things that are not solving their problems or don't provide the most value, but touching a little bit on the process and a B testing, really curious on your YC journey and maybe some of the most profound lessons or maybe conversations that you had in that experience that really maybe changes trajectory of not only your two companies, but specifically the one that went through the batch, but also in your personal thinking and your models and how you approach. Dirk-Jan van Veen (0:19:18) - So they're actually the same entity. So we just separated the brand of syncware and query verb because we still had some active engagements on the syncware side. We went through Y Combinator with the syncware idea. Why I think we were accepted was, well, why complaints are indexes heavily on the Founders. And I think the idea of making robots of different brands collaborate was big enough for them to say like, hey, this could be a really big company. Taylor Kenerson (0:20:03) - And some of the most profound lessons maybe, or conversations that you had through that Experience that really shifted or changed the way that you were thinking of Syncware and bringing it to market. What were some of those moments that really stick out in your Y Combinator experience that you can look back at and say, wow, that was maybe one of the bigger moments that really pivoted or changed or shifted the trajectory of where the company was going. Dirk-Jan van Veen (0:20:33) - So one of the first things that Y Combinator says when you join the batch is your company is a train wreck. We know it, don't worry about it. You have to tell us and then we can fix it. And so Y Combinator is very much about minimizing big mistakes that you're making in your company. So, for example, before we joined Y Combinator, we didn't have a proper bi weekly product development cycle, and they taught us about that. And since then, our product development has been much more streamlined. But they give a lot of specific startup advice, but they will not tell you anything about your industry. So those insights have to come from customer conversations. And I think that's where you can say why Combinator has a positive impact on most companies is that they make you focus, or they emphasize that you need to focus on talking to your customers and writing code and make that iteration loop as fast as possible. So I do think we sped up that iteration loop when we went through Y combinator, and that has stayed in our DNA, I would say that's really interesting. Taylor Kenerson (0:22:04) - And to your point, though, it definitely makes sense. I mean, focus is a superpower, and you only have 24 hours in a day. So what you do with those 24 hours is what your output is. And it is super important to only focus on what the customers are saying, but also on knowing your product, developing that, and having your intuition, like you mentioned earlier, and going back to that, and being able to rest and trust that, which is really huge. Just one last question before we wrap. What advice would you give to founders or anyone starting their journey in this entrepreneur space? It's really difficult. It's challenging, but it's also very rewarding. So what kind of tidbits or little wisdom would you drop to maybe your younger self, or a founder or an entrepreneur who is just about to embark on this journey? Dirk-Jan van Veen (0:23:44) - I think there's not really any very good generalized advice that isn't a platitude. It really depends on the founder and where they're coming from. What I would advise, but I think it's important to do it with a co founder. I think it will just make it more fun. So if you're successful, you can share that success with someone, and when things are rough, you can also share the grief with someone or share the burden. So I would recommend that in general, and all the other advice is, yeah, focus, iterate fast, maximize your rate of learning. Taylor Kenerson (0:24:39) - I love that. Dirk, thank you so much. I think that's really important to the partnership. And budy think, you know, life definitely can be lonely sometimes. And the founder journey is very lonely, especially when you have family and friends that are not very familiar with it. So having a co founder or a buddy that understands and has been on the journey and is walking alongside you is very important and can be very helpful. But also, on the flip side of things, maybe it's not ideal for some people. So like you said, there's no general advice, and to kind of just follow your own guide. Dirk, thank you so very much for joining us. Dirk-Jan van Veen (0:25:16) - Absolutely. Thank you. Adil Saleh (0:25:20) - Thank you, Dirk, for your time and insights. I love the way you've been processing and thinking a lot and giving a real nail biter answer, so thank you very much. Taylor Kenerson (0:25:33) - Thank you, Dirk. Dirk-Jan van Veen (0:25:34) - Thank you. Yeah. I hope it was good. Taylor Kenerson (0:25:38) - Yeah, I loved it. Thank you so much. Have a beautiful day. Dirk-Jan van Veen (0:25:43) - Well, thank you.