About
Building software since 2008, a professional platform engineer for the last decade, now going AI-first — I ship production LLM software and the agent tooling that builds it, across six domains at once.
The short version
I have been building for the web since 2008, and a professional platform and full-stack engineer for about a decade of that — multi-tenant platform software, release engineering, and the infrastructure around it (PHP/Laravel, Postgres, Docker, Kubernetes/Helm, CI). That is the day job, and it is real engineering at scale.
What this site is about is the other half: I build. Not toy demos — production software that ships and gets used, in more domains than any one role would ever ask for. The point of the breadth is not that I dabble; it is that the same engineering discipline transfers, and that I would rather learn a domain by shipping in it than by reading about it.
Six domains, on purpose
The projects on this site are not a single specialism dressed up five ways. They are six genuinely different problems, each taken far enough to be real:
- AI / LLM engineering — a production concierge product built on the Claude API: per-property knowledge base into a system prompt, server-enforced output scope, defense-in-depth guardrails, prompt caching and an answer cache that serves repeat questions in ~2 ms for zero tokens. Live, measured, multi-tenant.
- Games — Godot on Android TV: an arcade shooter that turns a phone into a gyro gun over UDP, and a two-player co-op platformer whose magnetic-charge physics constants were re-derived from measurement and regression-tested by a headless bot in CI.
- Mobile — React Native and Expo: a hardware remote with a full simulator so the stack runs without the car, and a hands-free Muay Thai trainer driven by voice routines.
- IoT / hardware — an eight-year RC-car project spanning app → WebSocket bridge → Arduino, with a motion-lease safety invariant that brakes on a lost link.
- Web platform & DevOps engineering — a productized site-build system with a Docker preview gate, CI validation and snapshot-plus-rollback production deploys; this very portfolio is one of its descendants.
- Agent automation — the practice that builds everything else (its own page).
The proof strip at the top of this page counts only what is actually published here — it can never claim a domain the work does not back up.
How I work
A few things are constant across all of it. Correctness is demonstrated, not asserted — headless physics tests, a full hardware simulator, measured latency budgets, health-checked deploys that roll themselves back. The safe failure is designed in — a parse that cannot be trusted degrades to one pack and a human check rather than a confident wrong order; a lost radio link brakes the car. And the plan is the source of truth: I write the intent down, keep it honest against reality, and treat drift between doc and code as a bug.
That last habit is why I have leaned hard into agent-assisted engineering — it rewards exactly this discipline. There is a whole page on it.
What I am looking for
An AI-first engineering role where building with LLMs and agents is the job, not a side quest — somewhere the bar is production software, measured, that real people use. I write in PHP, TypeScript/JavaScript, Python, GDScript and Bash, document in English, and I am based in Slovenia.
If that sounds like your team, the contact page has the two links you need.