Freelance IT developer — 18 years in IT (since 2008) · Linux systems · CLI tooling · local LLM / AI inference infrastructure. Self-taught, open-source first: from PC repair, networks and PHP websites (2008) to Linux servers and GPU inference today. I build small, fast, reliable tools and run real inference infrastructure on real hardware.
📍 Sidi Bouzid, Tunisia · 🌍 Open to relocation (EU / Germany / Canada) and remote roles 🛂 Non-EU — eligible for Germany's IT-specialist route (§19c AufenthG / EU Blue Card experience pathway): 18 years' IT experience, no degree required. ✉️ iamhsouna@gmail.com · 🤗 huggingface.co/iamhsouna · 💼 linkedin.com/in/hsouna-zinoubi
- Local LLM inference — llama.cpp, vLLM, GGUF model conversion (published on Hugging Face), GPU memory & model-compatibility troubleshooting.
- Linux & systems — multi-machine infrastructure, Docker, Proxmox, server administration, observability.
- Developer tooling — fast, focused CLIs in Go, Rust and Python.
| Project | What it is | Stack |
|---|---|---|
| lintop | Real-time Linux system monitor with NVIDIA/AMD/Intel GPU support (an htop + nvtop in one) | Go |
| hfd | Fast, resumable Hugging Face downloader — multi-connection, SHA-256 verified, TUI | Rust |
| llamacpp-manager | Swiss-army CLI for llama.cpp: builds, presets, and the running server | Rust |
| lyra | Local song generation studio (TUI + web UI) over a GPU-backed runner | Rust |
| bouzid-fm | Self-hosted AI radio station: generation, TTS, AzuraCast broadcast, terminal control panel | Python |
| GGUF conversions | Published, documented GGUF model variants with quantization notes | Hugging Face |
Arabic (native) · French (professional) · English (fluent) · Italian (basic) · German (learning now)
Everything above is public and runnable — that's the point. If you're hiring for inference infra, platform/SRE, or developer tooling and can support relocation or hire remote, let's talk.