Atomico Presents: Coffee with DeepL’s Jarek Kutylowski

Yeah, I mean, you know, in 2017, that was like really this super cool moment when neural networks AI that was not really yet fully known to the public. But if you've been watching what's happening in academics, if you've been into the maths of it and what was happening back then, that was this moment where we felt and I think a lot of people felt already like there's something bigger on the horizon and there's going to be a technology change coming up. And we've been playing around with language quite a bit already at this point in time and it really felt natural to start thinking about how will neural networks affect that. And then it was a little bit of a spur of the moment. Hey, let's just launch the service. Let's just see how it's going to work. We haven't been actually thinking much about how are we going to monetize that, how are we going to build a company really that's maybe sustainable, that's maybe a lot more long-term out of that. It was just this hey, there's technology, there is a problem that we know about. I'm born in Poland. I've been spending a lot of my time in Germany. I obviously have been learning other languages at school. So I think we knew what the problem is about and we've seen hey, there's this technology that can change how this field is going to look like so much. And obviously there's been large players in this field like all of the big tech. But we felt like this is the moment where they are actually going to have to throw away everything that they have for this technology and also start from scratch like we do. So why not?

So I've studied at a very theoretical university in Poland and I think Eastern Europe in general is kind of known for the fact that we can do the theoretical parts very very well. Maybe the practical applications of that and maybe the entrepreneurship at least at the point in time when I was studying wasn't so much there. This education itself was really good and it set the base for AI is an incredibly complex topic and the research behind that is a complex one. We've learned nowadays to use LLMs as just hey, let's just prompt it but there's a lot behind that technology and if you want to build your own models, if you want to design the architectures for them, you have to understand all of that background that's going on there. And I think the kind of theoretical aspect of that has helped me but I think it was not a common way for any of my friends, for anyone who was studying at my university or even then later when I did my PhD in Germany in the small city of Paderborn. There wasn't so many role models for that and I think that's one of those parts which have been really helping in the US, in the Silicon Valley where you just go out of the university and you see all of your friends going to other startups, funding their own startups. Some of those work, some of those do not work. And I think it's great if people know those options that are there and I think Europe has gone a really long way when it comes to that. There's so many cool companies that have been started over the last decade. I hope that's going to help the next generations to go on that journey.

Yeah, I think Europe is still very dispersed and there's not one central hub. So I think this ecosystem is not as tight as it might be. I think it would be a little bit easier for all of us if we were all living in one city or around one campus. But there's just now so many conferences. There's so many meetups. I think if there was something that I would wish for is a little bit more visibility for the ecosystem being driven by the governments and by the public sector really in Europe. That's something that I would like to see more of.

Yeah, I think the most important one that I really see is the visibility for every of our prime ministers, presidents say out loudly technology is what is going to be shaping our economies in the future and Europe needs to participate in that. I think that's incredibly important. The other part is I think what everybody's always talking about is regulation. Regulation is hindering businesses always in some kind of ways. If we're looking right now at the EU AI Act, it is a question for each and every AI company, but also not for AI companies like ours who are maybe really good at handling that. This is part of our business. This is at our core, but even for our customers who are then also less willing to adopt AI technologies and therefore where this market doesn't develop so quickly and so easily as it might be happening outside of Europe.

I think the biggest responsibility that I see on the ecosystem is to really move fast and build those companies and try to also do those hard things that are out there. There's different ways of approaching AI and there's different ways of building AI companies. We've been always very hard on the research side. We've been combining that also with the application layer and part of the deep thesis is that we can do the product, we can understand the customer but we can also build the AI for that. And I would really love to see more companies like that and companies who are thinking about how do we actually solve the underlying technology and who also really actively engage into that race with the big players outside of Europe. I think ethics is an important part of that. But I think in order to get into the discussion on ethics and how AI can meaningfully change our lives and how it can be there for the good and not for the bad, we first have to have those solutions. Otherwise we can't even engage into that conversation really. So the first step is to build.

Yeah, I think trust is important and trust is going to be even more important in the future as AI expands its capabilities. The quicker you let AI do tasks on its own, the more agency you give it, the more freedom you give it, the more the trust of you and myself as a human towards the technology is going to become more important. Right now in most use cases that we deal with AI, we're still the supervising party in some way. We're still having the checks and bounds on it. And I think in the coming years, this is going to be shifting towards AI going out there and doing things on its own. And for that we need to make sure that we both develop the technological ramifications on how can we actually limit it but also in a way that doesn't restrict its capability too much. That's an important part and I think in Europe we do have this mindset of thinking a little bit further when it comes to trust, to the consequences and implications of what technology can do. So doing that wisely, I think we can build great companies and great technology on top of that. And I think DeepL is a good example of that. We're being trusted by the most regulated industries. We are in healthcare, we are in the banking sector. This is one of our big advantages and I think for us, we're going to be building on top of that. But I think the whole ecosystem could leverage that strength too.

Yeah, I mean this is probably the toughest part about running a growing company and a company that comes to this size which DeepL is at is the learning part because you really have to change yourself each and every year. We've just gone through a larger offsite with the leadership team of DeepL and one of the slides that were presented by our coaches at the beginning of this session have been actually a whole slide of where Yaric does need to improve, what does maybe work but what doesn't work right now. And I think you have to have a lot of fun in learning and growing yourself and not getting too panicked about all of the feedback that you're getting because you're going to get a lot of negative feedback because you're going to be making a lot of mistakes there and you have to be just calm about that and you have to go on and move forward. It's at times really really hard but it is also a lot of fun and if you look back at those years and think what as a founder or as a CEO you've learned on this journey and what you can do now and what you could do like maybe even 12 or 18 months ago, it is really sometimes very very fascinating.

Yeah, I think this balance is super important because you have to distribute responsibility but part of your prerogative also as being a founder and not only CEO, I think also one of your jobs and duties and obligations is to go a little bit beyond that and be also in the detail here and there and also making sure that you're keeping a close watch on what's happening within the company so that you're maybe even the first one who notices rather than this information having to percolate through the company. It's a lot about giving responsibility, assigning trust. You need to have the right people for that and that's probably the most important job that you really have.

I think it's probably maybe even like 50% of my time that goes into that. And it is a broad topic because it's not only like directly doing that interview or doing the salary negotiations which are obviously understood as part of the process but really even thinking about what is it that I need from my leadership team today? What is it that I need from my leadership team tomorrow? How do the people that are currently there, how do they fit into that? It is critical. Like honestly if you would have asked me seven years ago I would say yeah that's important but I wouldn't have that conviction behind that. By now I have seen what a difference it makes to have the right person at the right place in the company. I think seeing people and how they are growing that's really very fulfilling.

I think it's been changing with the company and also it's been changing with even myself as a person and we've just talked quite a bit about how that's a growth journey. I think we've been very local, we've been founded in this small city of Cologne in Germany and have been consecutively over the years growing more and more into other parts of Germany, other parts of Europe, other parts of the world by now. We have offices in Japan, we have offices in the US. And I think what we've been trying to make sure that always stays is this open and trusted environment where any kind of problem can be brought up, where the best opinion wins and where there is this kind of trust of sharing ideas and listening. On the other hand, I think the culture has to also evolve as the company becomes bigger. Even if you're looking at performance in a company that is growing, there's very many things that can happen implicitly in a small company that you have to make explicit at some point in time.

I think it's an ongoing journey and I think when you maybe pass like 200 people in the company it starts becoming more and more important. I think we still have to catch up on quite many parts of that. I mean we've been growing fast. It's not all perfect. And even the evolution of the market like AI is like we are in the middle of this AI whirlwind tornado. That changes the culture of the company too. We have to become faster. We have to become bolder about what we're doing. Both internally in how we develop things, but also externally in how we present it just because this market is so full of messaging and news coming out from so many different companies. So there's also cultural changes that are going to be imposed by external factors to any company.

Yeah, I think the big change that was and it's still there on the table for DeepL is really figuring out how we can power more sophisticated and more complicated use cases in larger businesses in enterprises and how does that tie into the kind of core AI research that we're doing? How can we enable our teams to understand what the problems are in those businesses? And usually that's pretty complicated use cases. It's pretty complicated problems. It's not that easy of a horizontal consumer level product anymore. And how can we build the AI for that? And as I said earlier, the thesis around DeepL is really to bring it all from the customer down to the core level AI research. And we're trying to make sure that we find out those ways on how this can happen in a scalable way. How we can have all of our teams kind of being able to go out to the customer, understand the problem, build the AI tech that's necessary for that in a kind of as simple way as possible, but still solving the problem. And I think that's a lot of what's happening also out there in the world. Like we have this amazing AI technology right now, but very often we don't even know how to use it in practice. And I think DeepL needs to be really excellent at doing that. We've been really good at combining the product side and the research side. We have to get to another level there. And we've been always kind of on the outlook and always looking into our capital allocation and in our resourcing in tech. What is the next products that we're going to be building? We see ourselves as a company that's always innovating and I think if you're in AI, you have to have that spirit. You always got to make sure that you think about what is your next product. We're actually going out with some alpha products right now to our customers and testing them internally. I can't speak about those right now. You're going to hear about that probably in the next few months.

It's definitely different. The way that businesses or enterprises in general think about DeepL right now is a different one. A few years ago we've been a solution for this language problem for machine translation and everyone who was thinking about language obviously kind of came across DeepL. And I think if you're looking at language teams across the world they're going to be looking into DeepL but there's also many more of those strategic AI projects happening in enterprises. It's basically every board in every company asking the executive team like where are you employing AI in order to make this company more efficient? How are you making sure that we're staying technologically on top of things? And some of our customers engagement, some of our projects actually come from this space where the customers are looking for proven AI solutions, AI solutions that have been around for maybe a little bit longer than 12, 24 months and when there's a clear signal from the market that there is a ROI on deploying those.

Yeah, there is a lot of great technical talent and from Eastern Europe where I think stereotypically as I said there's a little bit more even theoretical knowledge there, swiping over to Western Europe where the education system is maybe a bit more practical. There's humble talent which wants to build yes but it's maybe also not particularly ego-driven which for companies like ours where the growth mindset and learning every day is super important. And that really works very well. I think where we have to step up a little bit is also teaching ourselves that we can be proud of what we can and be a little bit bolder and also be more willing to take risks. Sure, going to a corporate job is maybe a little bit safer, maybe a little bit easier, but the enjoyment of building something from scratch, that's also there. And that is what a lot of our employees really love about DeepL, but also what I know other startup employees are loving.