Optimization scientist in Strasbourg, France. I build planning and scheduling engines that run in production: rostering, timetabling, routing and blending. I work on MILP, constraint programming and metaheuristics, and lately on LLM agents that write and repair solver code.
Technical lead at Dots & Lines, an optimization and AI consultancy. PhD in AI and operations research (UTT, industrial thesis with Adesoft).
Current work
- Lead solver engineer on the shift design, rostering and sector assignment service for 100+ air-traffic controllers at Avinor (Java, Timefold, MIP, Kubernetes on Azure).
- VibeSolve: describe a planning problem in a paragraph, get a tested Timefold solver project. Agents write the model, then check and repair the code until it solves.
- Timestral in the browser: time series forecasting with Mistral's Timestral models, run locally with onnxruntime-web. Live demo.
Open source and research
- pyscheduling: Python framework for single machine, parallel machine, flow shop and job shop scheduling, with exact, CP and metaheuristic solvers (GECCO 2023 Companion).
- ITC2021: our entry to the International Timetabling Competition 2021 (sports scheduling).
- Papers at GECCO 2022, OLA 2023, META 2021 and CPAIOR 2022 on ML-guided scheduling and metaheuristic tuning. Google Scholar.
Tools I use most
- Solvers: CPLEX, CP Optimizer, Gurobi, OR-Tools (CP-SAT), Timefold, Hexaly, HiGHS
- Code: Python, Java, SQL, TypeScript
- Delivery: FastAPI, Docker, Kubernetes, Azure
Talks and awards
Optimization Grand Challenge award (2026). Speaker at Optimization4All 2026 on MIP versus metaheuristics for scheduling. Former lead organizer of GDG Algiers.
Website · LinkedIn · Google Scholar
Open to optimization projects, retainers and advisory work.