The German-Canadian startup yasp, founded in 2025 by Stefan Krassin, focuses on an intelligent, hardware-independent solution to significantly accelerate AI training and inference. The team recently raised $5 million, including from Cologne-based investor Capnamic. In this interview, co-founder and CEO Stefan talks about how yasp came to be and what challenges they are facing right now.

How would you explain yasp to your grandmother?

Imagine you are driving your car. But every road is different: sometimes it’s smooth, sometimes bumpy, sometimes paved. To reach your destination safely and quickly, you have to constantly adjust your vehicle’s settings as well as your driving style. yasp is like an automatic driving assistant: it helps you and your car to automatically make the right settings and adjustments. Whether you’re driving on the highway, off-road, or in the mountains, your car will perform in the best possible way. In our reality, the road is the hardware, and the car is the AI model. Currently, it has to be constantly retrained and manually adjusted for each specific purpose in order to make the car faster, more comfortable and better.

How do you intend to generate revenue, and how exactly does your business model work?

We’re a deep tech AI company, and we utilize a subscription model where our customers pay flexibly on a monthly or annual basis to access our technology.

How did you come up with the idea for yasp?

It’s pretty simple: we are solving a specific problem that’s affecting more and more companies. Artificial intelligence requires a lot of resources, especially when it comes to development, training and application – or “inference” in tech speak. That’s where code optimization becomes a game changer. Right now, NVIDIA offers the best overall package of hardware and development environment. However, this leads to the already scarce expertise in code optimization becoming even more tightly concentrated on this platform. The result is a lock-in effect for companies, causing them to lose flexibility. And that’s exactly where we come in, automating code optimization for any hardware with our Agentic AI Compiler. For developers, this means less focus on tools and infrastructure, and more time for their core task: developing AI models.

So far, what challenges have been the hardest to deal with?

One of the biggest challenges is the fact that we have developed a technically very complex product. So far, even OpenAI & Co. have not ventured into this area. We are pioneering this field and are creating a completely new category. At the same time, we are operating in a constantly changing market environment; technological developments in the field of AI are advancing extremely fast. Especially in terms of competition, we have to keep innovating and continuously set ourselves apart.

Which project is at the top of your agenda right now?

The official launch of our Agentic AI Compiler! Initial tests demonstrated significant improvements in performance for various training and inference workloads – without compromising on performance or accuracy. Now it’s a matter of getting it ready for market as quickly as possible. We’re currently in the final phase and will be launching our product in a few weeks. The waitlist is open and already quite long.

You have successfully raised investment capital. How did you establish contact with your investors?

Our technology has tremendous potential. We are confident that it’ll grow into a fundamental technology in the upcoming years. Fortunately, this has placed us in a very comfortable position, and we were able to close a significantly oversubscribed seed round. Still, investor demand remains huge.

Looking ahead to the next year, where will yasp be?

The next year will be crucial. We want to expand our customer base and further develop our product so that it meets the market needs even better. At the same time, we are expanding our team at our offices in Munich and Montreal, and are focusing on our target markets in Europe and North America. We are planning to open a third office in the US by the end of the year. Our goal is to grow strongly but sustainably, and we expect to start fundraising for our Series A round in the middle of next year.