Stefy L.

I build and self-host AI infrastructure on Linux, and I make sure it keeps running. 25+ years of UNIX/Linux systems engineering. I co-founded Devuan, the systemd-free Debian fork used worldwide, and authored Amprolla, its repository overlay system. That work is public and verifiable. WHAT I DO MOST OFTEN - Self-hosted LLM deployment. Ollama, vLLM or llama.cpp behind an OpenAI-compatible API on your own server, so your existing code works unchanged and nothing is sent to an external model provider. NVIDIA, AMD, Intel, or CPU-only. - Private RAG and document Q&A. Retrieval systems that answer only from your own documents and cite the source passage for every claim. Qdrant, Chroma or pgvector, local or hosted embeddings, OCR for scanned PDFs. I built and operate a production legal-document retrieval platform for an Italian law firm. - Linux server setup and hardening. Firewall, key-only SSH, fail2ban, TLS, automated backups with tested restores, monitoring, and a written runbook. Ubuntu, Debian, Devuan, RHEL-family; any provider or bare metal. ALSO IN DAILY USE Python, Bash, PHP, C, C++, Go, SQL, MQTT. Docker and Podman. Multi-tenant hosting with strict per-tenant isolation. Mail infrastructure: Postfix, dbmail, carrier-scale messaging. 40+ MCP servers behind an OAuth 2.1 gateway I wrote and operate, publicly listed in the main MCP server directory. A multi-provider AI broker published on PyPI. Reverse engineering of undocumented binary formats. HOW I WORK I am at my best with small businesses and offices that need one technical person who actually answers, and who owns the outcome rather than handing back a ticket. I will tell you honestly when something will not work, or when the problem you are describing is not the problem you actually have. Italian native speaker, so I can work directly in Italian as well as English.