Before
I didn't grow up thinking I'd work in AI. I grew up thinking I'd make things. Take apart, understand, put back together — it started very early, and it's still the thread.
Preparatory classes, then a software engineering school (ENSSAT, 2005). Then nearly ten years at IT services firms writing Java for major accounts — Teamlog, then Thales Services, on assignments for Orange, La Poste, the SNCF, Bureau Veritas, MACIF. I learned to ship under constraint, for systems that aren't allowed to go down.
The turning point came in 2014: Scrum Master at Conserto, on a long assignment with France's national health insurance. I liked the work enough that Cnam hired me in-house in 2018 — where I became a product and technical lead. Mission-critical environments as a school of rigor — it's still my day job.
The turn
In late 2025, I felt the line between “the one who imagines” and “the one who codes” turning porous. The tools — you know which ones — were collapsing the cost of experimentation. I made a simple decision: open a workshop, quit nothing, and build.
Not to chase a unicorn. To make objects that are worth something in themselves. Small, right, finished. The engineer's craft carried into the LLM era — run alongside twenty years in mission-critical systems.
Today
I run Tako, a platform that brings conversational AI back to where it's useful: the everyday life of freelancers and small teams. And Lima, a PWA built for theater improvisers — a product that didn't exist and deserved to.
In between, I take on a few client projects directly, no platform, no commission. Short, dense contracts where I work with the client — not for them.