Side projects
I’m very focused on respecting my personal data, my digital footprint, IoT, self-hosted, and in general on new tools and software: I always have a few projects to keep busy, learn, or automate my daily life.
Technologies
Here’s an overview of the technologies and domains explored in my side projects:
| Domain | Tech / Tools |
|---|---|
| IoT / Edge |  Rust, state machines, embedded Linux, local LLama, MQTT |
| Local game (grid-based) |  Typescript, RxDB, Yjs, Cloudflare Workers (Durable Objects/D1 & Pages), Auth0 |
| Self-host & personal infra | WireGuard / Tailscale, Ubuntu Server & RAID, Docker, monitoring (Grafana), TanStack React (personal projects), PWA frontends (ex: alternative frontend for lequipe.fr) |
Projects / Examples
- A set of experimental IoT tools in Rust for local devices and MQTT communications.
- A small local-first grid-based game (POC) using RxDB + Yjs for offline collaboration and Cloudflare Durable Objects for server logic.
- Several self-host projects: VPN, Ubuntu server, personal storage solutions, and alternative frontends (ex: https://github.com/the-boj/sport-results).
Freelance
I also take on occasional freelance work, mostly short engagements. Lately that’s centered on enterprise transformation through automation: identifying recurring processes, turning them into automated flows to improve reliability, reduce manual load, and speed up how teams work.
Teramed, Mindflow, QuickTask
Three engagements that read like a short timeline for how AI shows up in my work.
- Teramed — My last non-AI project
- Mindflow — The real transition to agentic coding at scale: integrated agents, more than just autocomplete or an occasional copilot.
- QuickTask — The first project where AI matters a great deal: spec-driven development, it writes most of the frontend code, supports me on the backend, and can sometimes run autonomous refactors on non-sensitive topics—it stays nowhere near anything infrastructure-related.
Experience feedback
I systematically try to test a technology that intrigues me: either because I have an idea, or to understand its scope and uses. It’s rare that I push a project beyond a small POC, except when it becomes useful on a daily basis (ex: my self-host stack). I stay on the lookout for new things and discoveries.
I’m strict about network hygiene, but that’s where I find calm: no social networks, no mainstream consumer messaging, almost no sign-ups on websites, clearing traces on search engines, and a dedicated email alias for each use. To keep everyday life workable, I’ve had to build systems at home. I run a small homelab—three or four machines and a router with outbound VPN: one hosts the apps (music, photos, Home Assistant, personal files), another handles DNS, inbound VPN, Tailscale, and my AI agent, and another runs a small AI model plus a few more services. That lets me stay continuously connected to the homelab from my phone and reach the same services most people use, but with gentler tools and more responsible handling of my data, on my own hardware.
On the personal-project side, I treat that agent as a day-to-day counterpart: Hermes Agent on OpenRouter, Firecrawl for web runs and synthesis when the answer isn’t in the weights alone, Linear to keep knowledge and shape issues, then Cursor from a ticket when it’s time to ship code. I use it for deep dives across unrelated topics, to drive automations and recurring chores, and what I’m really chasing is an agentic workflow that lasts for non-work projects—as much end-to-end autonomy as I can get from research and planning through implementation, with minimal manual stitching.
Thank you for reading!