I turn messy scheduling, routing and logistics problems into solvers that run in production. MILP, constraint programming, metaheuristics and machine learning, with CPLEX, Gurobi, OR-Tools, CP Optimizer, Hexaly and Timefold, from the first client workshop to the deployed API.
Each project shows how far it went: problem framing, mathematical model, implementation, production.
Optacare, 2026
When a nurse calls in sick, the roster has to be repaired fast without reshuffling everyone else. I designed and shipped the replanning engine of a hospital workforce-planning platform, built on a large-scale MILP.
Stack: Python, IBM CPLEX, watsonx.ai, FastAPI, Docker
Avinor via Dots & Lines, 2024–2026
Lead solver engineer on the production optimization service behind Avinor's workforce planning: demand-driven shift design, rostering and sector assignment for Norwegian air-traffic controllers.
Stack: Java, Timefold, MIP, Kubernetes, Azure
Logistics client via Dots & Lines, 2026
Buying, blending, trucking and selling crude oil in one plan. I designed a blending MILP coupled with a pickup-and-delivery vehicle routing model, linked by decomposition and re-optimized daily on a rolling horizon.
Methods: MILP, VRP, decomposition, rolling horizon
Open source, project lead, 2026
Describe a planning problem in a paragraph; get back a tested, ready-to-run solver project. A multi-agent LLM pipeline writes the model and constraints, then checks and repairs the code until it solves.
Stack: Python, LLM agents, Java, Docker. See it on GitHub
Recruitment client via Dots & Lines, project lead
Hiring panels need the right interviewers, a room and a slot that suits everyone. I led an application that builds interview schedules automatically, handles substitutes, and syncs with interviewers' calendars.
Stack: Java, Quarkus, Timefold, Google Calendar and Cal.com integrations
Adesoft, industrial PhD, 2020–2023
Automatic scheduling and resource allocation of rooms, teachers and equipment to courses and training sessions for universities and large companies. I built constraint programming and MILP models, then shipped a decomposition-based solver validated on real university data.
Stack: Python, IBM CP Optimizer, Gurobi, OR-Tools
What I reach for, and when.
MILP/MIP with CPLEX, Gurobi, HiGHS, CBC; constraint programming with CP Optimizer and OR-Tools CP-SAT; decomposition, warm starts, goal programming.
Timefold, Hexaly, local search, simulated annealing, genetic algorithms, LAHC, GRASP, ACO for problems too large to prove optimal.
Demand-driven shift design and rostering, machine and job-shop scheduling, timetabling, resource allocation, vehicle routing with time windows, blending and network flows.
Python, SQL (PostgreSQL, Supabase), pandas, scikit-learn, XGBoost, PyTorch; predictive modeling; ML for algorithm selection and tuning; KPI design and benchmarking.
FastAPI services, Docker, Kubernetes and Helm, Azure, IBM watsonx.ai, AWS, CI/CD with tests on every solver change.
Client workshops and spec interviews, proposals and presales, solver teams, training engineers on MIP, CP and metaheuristics.
PhD at Université de Technologie de Troyes on learning-based methods for parallel machine scheduling, the core of production planning in manufacturing supply chains. Full list on Google Scholar.
Explaining when to reach for which method is half the job.
Optimization4All workshop, 2026
A live, hands-on workshop on one scheduling problem solved two ways: an exact MIP model in Pyomo with HiGHS, then a metaheuristic in Timefold. It covers adding real-world rules, keeping scoring fast at scale, and what it takes to ship an optimizer to production. Colab notebooks let attendees follow along.
UTT, GDG Algiers, hackathons
Lecturer at Université de Technologie de Troyes (2020–2023) in C, Python and algorithms. Lead organizer of Google Developer Group Algiers for five years, with 200+ volunteers and 20+ talks of my own, and an organizer of the Hajj Hackathon, the Guinness World Record largest hackathon.
Optimization and AI challenges
Award winner at the Optimization Grand Challenge 2026. Third place at AI Hack Tunisia 2019. Competitor in the International Timetabling Competition 2021.
Open to lead optimization and data science roles in supply chain, logistics and workforce planning, remote or on contract. Based in Strasbourg, France. English, French and Arabic.