I’ve been building software systems for more than 25 years. Mostly backend-heavy, data-heavy, production systems. Some of them were used by millions of people, some were small internal tools, some were hardware devices that you could literally hold in your hand.
Over time, I noticed that I’m always drawn to the same kind of problems: systems that become messy as they grow. Systems where implicit assumptions pile up. Systems that start behaving like magic even though they’re just code. I tend to react to that by trying to make the structure more explicit — models, graphs, clear states, understandable flows.
Right now, most of my focus is on structured AI. Large language models are impressive, but they don’t really “know” anything. I’m experimenting with ways to build a structured knowledge layer around them: concept graphs, explicit relationships, simple inference rules. The goal is not to make AI sound smarter, but to make its behaviour easier to reason about.
The main project in that direction is Cogentis AI
It’s still a research prototype. I’m exploring how to represent knowledge explicitly and how to combine that with probabilistic models without turning everything into a black box.
Before that, I built URU, an open hardware FIDO2 security key.
It went through multiple PCB revisions, firmware iterations and a few failed prototypes before it became a real, working device. It was a good reminder that in hardware, there’s no room for hand-waving — either it works, or it doesn’t.
I also experiment with applied systems:
- Exploristo (https://exploristo.com, https://github.com/exploristo) — an attempt to model places and localities as a semantic graph instead of a flat list.
- DearHiringTeam (https://dearhiring.team, https://github.com/dearhiringteam) — a structured job application tracker, because hiring pipelines are usually opaque and chaotic.
- TelePico (https://telepico.com, https://github.com/telepico) — a small service for publishing images on the open web without feeds or algorithms deciding their fate.
Professionally, I’ve worked at companies like Acronis, trivago, Instana (later IBM) and TIER Mobility, building and scaling backend systems and, later on, leading engineering efforts. Before moving to Germany, I ran my own company in Prague and spent many years doing hands-on development in Moscow.
I’m comfortable working across layers — backend, infrastructure, data, sometimes frontend — because real systems don’t care much about neat boundaries. What matters to me is that the whole thing works and can be understood.
If you’re interested in structured AI, knowledge modelling, or just building systems that don’t collapse under their own complexity, feel free to reach out.