About

I’m currently working on mesh.applied, where we distill large open source models into highly capable domain experts for local inference. We have always been resource constrained, thus were forced to optimise everything in order to get a meaningful performance. We have traction, sold a complete inference unit with our kernels running on it. And we have a few other in the pipeline, the primary areas we’re focusing on are cyber, coding and research.

Previously to that I co-founded mesh. CEEs largest self-sustaining builder space in Budapest, with over 35 members now. We fund young talent so they can build whatever technology they want. We contracted and delivered multiple hundreds of thousands of USD worth of software engineering, primarily deeplearning solutions and prediction systems based on physical/mathematical simulations. In addition to that, we’ve organised 50+ events and have 3 spin off startups. I personally mentored over 30 people and interviewed over 200.

At mesh., I’ve met the founder of Resoloon, which pretty soon became one of my dream projects, with endless possibilities of engineering and research. In addition to helping out where we could, we ended up angel investing into the startup, later he went onto raising a pre-seed round in October 2025.

My building experience in tech started with 3, at that time, to me random people in the dorms kitchen, who later became some of my best friends. We’ve built many-many projects over the course of a few years, sometimes just for fun, sometimes because we ran out of money, and sometimes because we thought we would change the world. I should probably keep a list of how many things we’ve done, but it’s easier to just throw in a few links: Matteo, Petya and me.

My primary interests are deeplearning, reasoning, continual learning, and having as much compute on tailscale as possible.