Auto-Mail
An LLM sorts my mail, and what survives lands in a Tinder-style app on my phone.
Open to work, starting as soon as Fall 2026
Compilers, rendering engines, anything where you can follow a value from the surface down to the instruction that moves it — that is what pulled me in.
In practice, most of what I build removes work that repeats. A vehicle-fleet site rebuilt so a team stops maintaining spreadsheets. A company site handed over on templates its own staff can edit. Sixty thousand simulated trades to check, with numbers, that none of 22 entry techniques beat random buys. An inbox scored and triaged before I open it.
That last one is where my attention is now: putting LLMs to work on the tasks that pile up — dispatching them, wiring them together, and keeping the decisions they are not reliable enough to make in ordinary code. Auto-Mail is the working example; orchestrating several agents on one job is what I want to build next.
I am looking for a role where that work saves a company real time — configuring and integrating models, or writing the software that makes an existing process faster.
Available from Fall 2026, based between Paris and Aix-en-Provence and happy to relocate anywhere. I spent my exchange year in California, and the US is where I would most like to return.

An LLM sorts my mail, and what survives lands in a Tinder-style app on my phone.
Video editing through code, so you can vibe-code an edit with an AI.
A language, compiler and virtual machine written from scratch in Haskell, by five of us.
A macOS panel: every Claude Code session on my machine, which one is waiting on me, and the mail and todos that are waiting too.
Let's automate what repeats and scale what works.
I am available to discuss projects, collaborations or new opportunities.