Antonín Hoskovec
Tonda is an ML engineer with a PhD in mathematical physics, and as he puts it himself: "I'm a tinkerer, and machine learning is a fantastic toy." He loves operating at the boundary between research and industry — he's tried it at Seznam, Rossum, QMiners, and GLAMI. He's currently part of the Miton AI Lab and organizes Miton AI Times.
How would you describe your role at Miton?
I'm part of the Miton AI Lab, where we experiment with what technology has to offer business. I've established that my strengths are: I read technical papers quickly and happily, I maintain as broad an overview as possible across research and software, and know how to apply them in commercial settings; I'm not afraid of unfamiliar codebases — and I enjoy talking with people I can learn a lot from, including non-technical ones. I'd like to leverage both of those things, whether to bring in new investment opportunities or to review existing ones. Within the AI lab, I'm also planning to create a number of them myself.
What technology fascinates you most right now?
More than a specific technology, it's a field — understanding LLMs. I feel like LLMs have run laps around us in terms of how well they work, compared to how well we understand them. Being able to see inside their properties should improve our estimates of their limits, of what's realistically buildable on top of them, and so on.
What kinds of projects interest you most — by sector or approach? Who should come find you?
By nature, I'm a technical and thorough person with an academic background, so anything that requires a computer, a serious scale, or mathematics interests me. I enjoy e-commerce and also like helping with medical projects.
Tell us more about your academic background.
My PhD is in mathematical physics from FNSPE CTU in Prague, where I'm a member of the Q3 research group led by Prof. Jex. There, we work on quantum computing. My publications fall into three areas: machine learning, quantum computing, and medicine. But I've never been primarily focused on publishing — I'm more interested in practice. That's also why I now organize more programming work than research within Q3.
What's your tech stack?
I always come back to Python, but other languages aren't foreign to me either. Surprisingly, I spend quite a bit of time on devops and Kubernetes. I code through agents, of course — right now I have Pi everywhere.
What do you do with AI when you're not working?
I see a mismatch between how AI is actually used — and how people want to use it — for psychotherapy, versus what happens in practice. So I'm looking for projects that do it better. I'm also automating our housing association.
How did you end up organizing Miton AI Times, which over the years has built a strong community of technical people?
It's not just the last few years when new AI/ML technologies have been mushrooming like crazy — though now it's extreme, in machine learning it's been this way for a long time. You simply can't keep up with everything on your own, and hearing other people share their intuitions about projects you'd never stumble into yourself is incredibly valuable. So we went for it.
How would you describe your working style?
Enthusiasm → overwork → decent result. And a good vibe with the team.
Where do you spend most of your working time?
At the Miton offices.
What book do you give away most often?
Flowers for Algernon by Daniel Keyes. It's my favorite book, and I often find out someone hasn't read it.
What are the five things you can't live without?
Am I supposed to mention a computer and a phone? (Audio)books, a bike, a frying pan.
What's your favorite place?
Day to day? Ladronka. For keeps? This restaurant in Lisbon.