Full-stack developer · Debrecen, Hungary
Two years building software professionally. Today I build internal web applications at the University of Debrecen, Faculty of Informatics; a monitoring system and a course management platform among them. Before that I co-founded Skillion, an AI EdTech product, and took it from idea to a live service.
It reached ~10,000 registered users and hundreds of paying subscribers during that time.
Right now I'm moving toward AI engineering: LLM integration is the part of the stack I want to own, not just consume.
Java 21 · Spring Boot 3.2 · MongoDB 7 · React 19
Answers "what can I build right now from the bricks I already own?" Owned sets are merged into one part inventory, then catalogue sets are scored against it and ranked by completion.
The hard part: quantities decide it, not membership. Three 2×4 bricks do not
cover a requirement for eight, so every part contributes min(owned, required).
The upstream API allows one request per second, so the client paces itself, and
the paging calls go through a retry wrapper that honours the server's
Retry-After interval instead of guessing a backoff. Part lists are cached per
set, which keeps one page of results from costing hundreds of calls.
Java 21 · Spring Boot · React · PostgreSQL · Docker
Task manager with JWT authentication, role-based access control and an admin dashboard.
The hard part: the admin surface is guarded twice over. The URL pattern is
locked to the ADMIN role in the security config, and the controller carries its
own @PreAuthorize so the endpoints stay protected even if a route rule is ever
edited. Sessions are stateless, and role changes are checked against an allowlist
rather than written straight from the request body.
| Area | Tools |
|---|---|
| Languages | TypeScript · JavaScript · Java · Python · C# · SQL |
| Frontend | React · Next.js · React Native · Expo · Tailwind · Radix UI · Vite |
| Backend | Spring Boot · Flask · FastAPI · Node.js · Express · REST |
| Data | PostgreSQL · MongoDB · MySQL |
| AI | Claude Code · Cursor · MCP · LLM APIs |
| Delivery | Docker · GitHub Actions · Azure · Nginx · Maven · JUnit · Vitest |
I work AI-native day to day: Claude Code, Cursor and MCPs are part of the toolchain. Data Science MSc in progress at the University of Debrecen.
6 verified: NVIDIA, Azure ×2, Java, Python, Databases
Every badge links to its verification page.