The following interview is a conversation we had with Malte Pietsch, CTO & Co-Founder of Deepset, on our podcast Category Visionaries. You can view the full episode here: $45 Million Raised to Build the Future of LLMs

Malte Pietsch
Malte, thanks for chatting with me today. Yeah, hey, thanks for having me. 

Brett
No problem. Super excited to chat with you today, and I’d love to just begin with a conversation about what you’re building. So at a high level, what problem are you solving? 

Malte Pietsch
Yeah, I mean, as you know, we are kind of in the middle of a huge industrial transformation where AI gets really embedded into almost every product, every process, and it’s really, I think, becoming this new generation of intelligent applications and processes. And our mission at DeepSet is really to enable this transformation, enable engineering teams out there to really build, evaluate and ship those AI features they need for that product in a very fast and efficient way. And for that, we basically have two products. So one is haystack. So open source framework allows you to orchestrate applications into pipelines. It’s really targeting, say individual devs that build a single use case. And then there’s DeepSet cloud or SaaS platform where enterprise teams can really build end to end. 

Malte Pietsch
Iterate quickly over hundreds of different pipeline architectures, deploy that to production with one click, more or less, and observe then the quality replication over time. Yeah, that all. I think we really saw that the team spent a lot of time not always on the right things. What we really believe in is that they should spend time on customizing the behavior of the use case and not on any undifferentiated heavy lifting, how to scale efficient gpu’s, how to sync data from your sources. So that’s kind of what we abstract away for teams. 

Brett
From my conversations with other AI founders, they say that their company can really be divided into two general chapters, and that was pre GPT which came out in what like November 2022, and then after that. So lets focus on kind of that first phase or that first chapter before Chechpt came out into the world and everyone’s mother knows about chat gpt. What was that period like for the company? 

Malte Pietsch
Yeah, so we actually started quite a few years earlier, so were now more than six years in. And back then when we actually started, no one talked about llms, few talked about nap back then. And were in this early phase where it was with early signs that it comes easier to use these kind of transformer models. What is the foundation for something like GPT? And in the very early days of the company, then Google actually released their first model, first transformer model. And that, I think was really for us, an interesting flexion point. We jumped on that. I remember were in the middle of a customer project with Airbus back then. We jumped on a trained first model, tailored it to basically their domain knowledge and that I think really then determined our trajectory. 

Malte Pietsch
Worked a lot with open source companies, like hugging face them in these early days, and ultimately also launched our first open source product, Haystack. But, yeah, in contrast today, or like was it last year, I think after chat GPT, these first years were, I think, more niche. There was a small community. You could already solve pretty interesting use cases, but it was definitely more like a smaller scale, quite via target groups, data science teams, machine teams. And now I think it became more mainstream, which is amazing. Awareness just became bigger, easier and faster to build these applications. But it’s really now a new scale and a new awareness in the market. 

Brett
If we think about that period then from 2018 to 2022, were there ever any moments that you started to kind of lack conviction there of like, maybe there isn’t a market here or the whole time, did you know this is going to be a massive market. It’s just a question of when. 

Malte Pietsch
Honestly, I think we always have this conviction. Yes, I think for us was always, I don’t know, we had very strong beliefs from day one that it’s so obvious to us that at some point you would talk with machines, you will rather chat with products. We use language to steer controls in your car rather than clicking lengthy wizards and so on. So I think we always had this belief. It wasn’t, of course, not clear to us, like, okay, when is it going to happen, at what speed? And there were moments where we thought, okay, maybe it’s taking longer than we think. That definitely happened. But I think really like the key part for us, that were always very close with customers. So the first years we actually bootstrapped and definitely got us. 

Malte Pietsch
We were always very close to some feedback and was always very rewarding when even back then you could solve customer cases. Customers paid us for what we did. They saw some value in what were doing, what we helped them with NLP. So even though I think were not at that stage yet and not at the level of use case you can solve today, I think this close customer feedback and signal was always very encouraging. 

Brett
Are you surprised at the pace and the speed things are moving at today? 

Malte Pietsch
Honestly, kind of. I think last year was really astonishing. I think not so much that it happened, but rather than how I think the whole year went and how fast, I think the kind of markets jumped on it and how much I know how many new startups grew overnight, how many open source projects were started, how much customer interest there was from one day to the other. So you say internally with an off site recently where my Co-Founder described how you felt over the last year and was a bit like a washing machine, constantly gets turned around. You wake up in the morning, you check, oh, what’s happened now? Again, where you have to very, I think, quickly adjust to where the market is going, what customers want, we’ll get out and release what model overnight, what did, what may I do here? 

Malte Pietsch
What is, I don’t know, Google’s reaction to that. So I think that was very turbulent year for us. Very exciting, but also very, I think, fast and dynamic. 

Brett
I’m sure in this post chat GPT world you’re dealing with more competition and maybe not even direct competition, but more noise. There’s obviously a lot of funding that’s going into it. A lot of ecs are spotting the opportunity a lot of founders are building. What are you doing to rise above all of that noise? 

Malte Pietsch
Yeah. So I think it always depends with whom. Honestly, I think we are not the loudest company out there and probably not the best also in terms of just raising our voice and then catching attention, I think how we see it, nothing special. Now, about this year I think it’s really about bringing prototypes to production and showing clear business value, solving real use cases and not just building demos. Last year, I think that was where a lot of this noise happened, where so much new things became possible. So a lot of cool ideas, great ideas, exciting demos. But I think at some point, if you don’t want to become hype and just a bubble, you need to show the value coming out of things. 

Malte Pietsch
And that’s, I think, where we come from our bootstrapping history, working always with customers very closely, really bring stuff to production. And that’s, I think, also what we doubled down on this year, really trying to work closely with customers and I think end of the year we will see from a lot of people in the market, companies in the market who can make that step, who can really bring customers over this finish line and where will they best basically just stay in POC stage. 

Brett
How would you define then the marketing philosophy that you have? Is it just to build a great product and get it in front of the right folks or what is that marketing philosophy? 

Malte Pietsch
No, I mean we have of course a lot of things that we do on that side. And I think to one degree we have open source like Haystack which really gives us a lot of awareness and helps us actually to be known, being brand, being very transparent in our technology so people really know what they’re getting also if they decide to go for the commercial product. So I think that’s basically one whole theme where we have marketing, where we go to events where we publish a lot of content, especially they are targeting developers. And then of course on the other side we try this year to really get more or less like two messages across. One is really around this use case orientation. 

Malte Pietsch
So we know from a few of our customers that this in that use case is interesting that they saw that’s working for them. So we tried to double down on that and really focus there on messaging more product owners, business lines who maybe also have that problem. So a bit like solution oriented marketing there. And then similarly in parallel we have a few campaigns that rather focusing on platform because we also have some customers who rather come from the side. We know we have x use cases, some are already in production or advanced, but we need this one platform where unify our work. So it’s often rather large enterprises with some kind of platform team for them we have rather this say technical horizontal messaging. Why? Like what parts of our platform are really interesting for them? 

Malte Pietsch
How can that platform help them in their work? That’s the thing to main approaches. Then I think for our working market what we do is we sometimes call it sandwich motion. So I mentioned this open source product. It’s really a lot of developers out there know it, tried it, like it and typically like the devs are not our buyers. Like they don’t swipe the credit card and pay the check for this commercial platform, but they help a lot in this buying process. So when we enter a company, an account, have conversations often and with product owners, with business lines, what use cases they want to solve. At some point, of course they ask the developers about hey, what is that technology? Does it make sense? Do you trust us. 

Malte Pietsch
And that’s then I think we have a lot of credibility from the developers where they say, oh, yeah, it would be actually cool. We tried to source products. Amazing. To continue building with that, let’s go with the commercial one. But for us, it’s really, you need those two checks on the developer side, but then also usually on the buyer side, which is somewhere else. 

Brett
This show is brought to you by Front Lines Media podcast production studio that helps B2B founders launch, manage and grow their own podcast. Now, if you’re a Founder, you may be thinking, I don’t have time to host a podcast. I’ve got a company to build. Well, that’s exactly what we built our service to do. You show up and host and we handle literally everything else. To set up a call to discuss launching your own podcast, visit frontlines.io podcast. Now back today’s episode. 

Brett
At what point did you have a second product? Was that from day one that you had two different products, or at what stage in the company history did you expand to have a second product? 

Malte Pietsch
So it was not always the case. We started with open source in first game, we had bootstrapping. Then I think two or three years in, we launched haystack, the open source product. And that actually got adopted quite fast and quite well. So that was quite early. Clear, okay, this is going somewhere. Community was growing. Netflix was one of the early contributors and adopters and then kind of continued to grow. And then I think that was probably different to many other open source companies out there. We more or less from day one, then thought, okay, what’s the commercial product like? What’s the business model that can really create here to have a sustainable business? And I think for us, that was very important in that stage than to talk to other founders, other commercial open source founders in particular. 

Malte Pietsch
Then we talked with Shai Ben from Elastic, Andre Alexei, who did engine x and then became actually an employee. And our team, full employee, and also the databricks team, if we talk. And I think from that it was all clear that our way, how we wanted to do it, is not just grow the community massively and just focus on open source for a few years and then figure out later how was the business model for us. I think our way was rather, let’s do open source and on the sideline, try a few different things, launch a few prototypes, experiment a bit, basically take our time there to do these experiments. And then when we feel comfortable about a product, if we see some product market fit, then we basically grow that and really launch that. 

Malte Pietsch
And yeah, I think that basically then resulted in ViPsat cloud which we launched more or less like one and a half years ago, I think, and I think that’s happy with our approach to happy how both products are now developing from, I think from a model can think of it as the databricks approach, where Haystack is really the technology, the open source technology. But then hey, keepsake cloud is the full platform, a bit like databricks and Spark. 

Brett
To date, what do you think has been the most important go to market. 

Malte Pietsch
Decision that you’ve made? I think at some point we launched that company always with having in mind, okay, this will become a product company, you want to build a product. But in our early days certainly were bootstrapping because we felt, okay, we have these few product ideas, we didn’t want to place a bet from day one one of them. So we started basically doing professional services. We built custom AI solutions, g learning solutions for enterprise customers. And yeah, we started with, I think we had €5000 on the bank in the beginning and that’s how we operated and started. And we realized pretty quickly then that this professional service business is very profitable. So we grew the company very profitable to quite some revenue. But it was not, we always had and knew, okay, actually that’s not our goal, right. 

Malte Pietsch
We didn’t want to build a professional service company, but at some point we had them to make this decision and really think, okay, what kind of company do we want to build? And of course that’s then also what kind of go to market do we have here? Do we sell professional services or do we sell actually a product? And I think that was not a tough decision really. But okay, at some point we had to sit down and make it consciously. And yeah, we found us and said, okay, let’s go for product. Even though I think the other path would have been maybe easier from a financial perspective, for sure, also attractive. But I think we all were just intrigued by the etiquette here. We want to rebuild a product, we want to see how our technology runs all around the world. 

Malte Pietsch
Being really embedded in all these intelligent applications that we envisioned, really building some kind of legacy. 

Brett
I would say, as I mentioned there in the intro, you’ve raised over 45 million to date. What have you learned about fundraising throughout this journey? 

Malte Pietsch
I would say shouldn’t optimize for valuation, but always for partner. I think that’s what many people say, but it’s truth. And I think looking back, we always did that in the funding rounds and at some rounds it felt weird to some degree. We said, oh, we have these other offers, but we don’t go with them. F higher valuation, all cool. But we didn’t fear good vibes with a partner or didn’t really see how this collaboration is the best thing we can do. So yeah, we always think then rather optimized and picked the partners that we felt are helpful in that certain stage. Different partners? I think so also in every stage we picked and in the long run. Now, if I look back, I think that was definitely the right decision. Also, I think in terms of market, it’s a finger. 

Brett
It was a healthy decision outside of fundraising. Let’s imagine that you’re having a conversation with a early stage builder who’s building an AI product or something in the AI world. What would be your number one piece of advice to them based on everything that you’ve learned so far in building a company? 

Malte Pietsch
Be close to your customers. So I think for us, this bootstrapping phase was incredibly valuable because we just learned so much about really pain points of these customers in various companies before we had anything. We’re super flexible to really go into different problems, different projects, bring out different stuff, and I think that I wouldn’t miss that. I think it’s very risky if you are a tech company, if you build a very deep tech product that you may sometimes focus too long on just the technology because you have this vision, you have this idea, you just need to build it, but then maybe you hit the market too late and realize it’s too late. That okay, no, it’s actually we’re only these, I don’t know, two, three customers that had this problem that we talked to, but not the 200 others. 

Malte Pietsch
What’s not the biggest problem that you look for, that you build something for? There was something else that would have been even better. So I think I would really encourage people going all in on customers, even if it’s not, maybe not even the direct fit for your customer, for the product in early days, I think it’s really worth it. Maybe the second one would be retrospective from last one, two years. I would probably also hire a sales team earlier, to be honest. So we did, I think, for a very long time. Founder led sales and only hired last October. Someone said professionally for sales, and I think also that could have been earlier, I would do that next time differently. 

Brett
What did you learn from making that transition away from Founder led sales? I’ve seen it with a lot of founders. It’s very difficult to make that transition. 

Malte Pietsch
I think for us it was kind of a relief because back last year, I mean, it was really starting to became chaotic. So much was going on, so many customer requests, and we had, I think it was hard to entertain all of those conversations, to really be quick also with customers if they have requests. And I think now that we have their system in place, I see how things are developing in a very healthy way, very structured way that I think was fast. Not really, I would say a hard transition, but actually for us, what was hard? I think finding the right person, the right leader for sales. That took us actually longer than before. 

Brett
Final question for you, let’s zoom out into the future. So, three to five years from today, what’s the company look like? 

Malte Pietsch
Yeah, I think we are. In three to five years, I really believe AI will be embedded in all products, like in our daily life. All processes or products will have some AI feature embedded. And, yeah, hopefully a lot of them are powered by haystack, a lot of them are powered by cloud. I think for me, one of the biggest motivation factors and boosts is if I get out, talk to random people, and they say, oh yeah, I know haystack, we use it for this and that use case, and we build something here and there for our product. So I think in three to five years, I just want to have even more of these conversations all around the world. 

Malte Pietsch
And company wise, I think we will be global where this year or last year we already expanded to the US, but I think be probably even more present over there. And I think right now, a lot of big logos, interesting customers in the select this early stage in our pipeline. I open in a few years from now. We have many of them live, and many of them have exciting use cases that they launch. Amazing. 

Brett
I love the vision and I love this conversation. I know it’s almost 08:00 in the evening your time or past 08:00 in the evening your time on a Friday. So thank you for taking this end of week. I really appreciate it. It’s been a lot of fun. If there’s any founders that are listening in and want to follow along with your journey, where should they go? 

Malte Pietsch
Hi, you can find me on LinkedIn, on Twitter. Usually I’m more responsive. I mean, I benefited a lot, as I said, from other founders in our early days. So please just ping me and give me a message. I’m happy to help where I can. 

Brett
Amazing. 

Malte Pietsch
Thanks so much for taking the time. Thanks, Brett. 

Brett
This episode of Category Visionaries is brought to you by Front Lines Media. Silicon Valley’s leading podcast, production studio. If you’re a B2B Founder looking for help launching and growing your own podcast, visit frontlines.io podcast. And for the latest episode, search for Category Visionaries on your podcast platform of choice. Thanks for listening, and we’ll catch you on the next episode.