Engineer × Lawyer × CPA

I’m Michael Benesty. I write about what I learn at the intersection of code, doctrine, and day‑to‑day practice.

I’ve been a financial auditor in New York (Constantin Associates), a tax lawyer in Paris (Deloitte), and now an ML engineer. Speaking statute, ledger, and GPU is quite useful to turn legal reasoning into tools professionals actually trust.

By day, I lead R&D at Lefebvre group, European leader in legal and tax publishing, where we’re shaping GenIA‑L, a cited, editorially‑anchored assistant for legal and tax work. In open source, I helped build the official R package for XGBoost and co‑created tooling like Kernl and transformer‑deploy because useful AI based products should be performant and easy to ship. Earlier, with SupraLegem, I learned a lot about what responsible legal open data looks like (and what it doesn’t).

What you’ll find here: practical notes, design choices, and some experiments results.