I build AI agents, ship fast APIs, and reverse-engineer what's under the hood.
🎓 Economic Cybernetics, Statistics & Computer Science at ASE Bucharest (2024-2027)
Third-year student who builds AI agents and the FastAPI backends behind them, and who likes taking systems apart to see how they really work. Most of what I know I learned by shipping working tools under hackathon pressure: one problem, a deadline, and code that has to run at the demo.
- 🤖 AI agents: tool-calling pipelines, multi-turn context, web search and human-in-the-loop approvals (PrivAI, JetSeeker)
- ⚡ FastAPI backends: SSE streaming, webhooks, background tasks, local-first storage
- 🔍 Reverse engineering: mapping undocumented production APIs from network traffic, sessions and polling (JetSeeker × Vola.ro)
- 🔗 Real-world integrations: live government APIs, rate limits, caching and hardened parsing of untrusted input (InvoiceGuard × ANAF)
- 🔐 Security by default: PII redaction before prompts leave the device, XXE-safe XML, data stays local unless it has to leave
Right now: building with AI coding agents to go from problem to shipped tool fast, and digging into what makes agents safe to trust, the point where AI becomes both a powerful tool and an attack surface.
🛡️ InvoiceGuard |
|
🔐 PrivAIAI Agent Olympics, Milan AI Week 2026 Fiera Milano, Italy · May 13-20 · $32K+ prize pool Privacy-first AI workspace where all data stays on-device in SQLite. Built with a team of 4 at Europe's largest AI conference. Features a privacy guard for PII redaction, autonomous agent workflows, and Gemini-powered multi-model intelligence.
|
|
🇷🇴 Romanian (Native) · 🇬🇧 English (C2) · 🇫🇷 French (B1)
💼 Open to internships, hackathons, and collaborations in cybersecurity & AI security.