PhD candidate at TU Clausthal, finishing a dissertation on the Vehicle Platooning Problem: coordinating truck routes and schedules so vehicles form fuel-saving convoys on real road networks. I work on both sides of the methodology, exact optimization (MILP with Gurobi) and metaheuristics, and benchmark them against each other at scale. Beside the research, I have seven years of university teaching and thesis supervision.
I care about clean mathematical models, reproducible experiments, and code that keeps the problem apart from the algorithm.
- Exact optimization: MILP modeling, Benders decomposition, two-stage stochastic programming (SAA), chance constraints · Gurobi, CPLEX, HiGHS, SCIP
- Metaheuristics: GA, SA, PSO, ACO, CMA-ES; multi-objective: NSGA-II, MOPSO, SPEA2, PESA-II, MOEA/D, built as reusable, problem-agnostic engines
- Supply chain: inventory routing on real e-commerce orders, stochastic facility location, vehicle routing, shop scheduling
| Repository | Summary |
|---|---|
| two-echelon-inventory-routing | Warehouse, hubs and van routes on real Olist orders in Sao Paulo with OpenStreetMap road distances. Greedy rule, exact MILP (Gurobi), ALNS and a matheuristic. Gurobi proves the optimum on all 12 small instances and the ALNS reaches it in 120 of 120 runs. Paired Wilcoxon tests on the larger instances. |
| multi-objective-vehicle-platooning-problem | Joint route and speed selection for truck platooning. Six multi-objective metaheuristics benchmarked against the exact Pareto front of a MILP solved with Gurobi. The core problem of my PhD. |
| multi-objective-scheduling | Energy-aware tri-objective flow-shop & job-shop scheduling (makespan, tardiness, energy) on the Taillard benchmarks. NSGA-II, MOEA/D, and MOPSO from scratch with adaptive operator selection, Dockerized PostgreSQL experiment tracking, and a full nonparametric statistical comparison. |
| stochastic-facility-location | Two-stage stochastic capacitated facility location with SAA, a service-level chance constraint, and Benders decomposition. Solver-agnostic backends (HiGHS / SCIP / Gurobi). |
| multi-objective-optimization | NSGA-II, MOPSO, SPEA2, PESA-II, and MOEA/D implemented from scratch and compared on benchmark problems. |
| genetic-algorithm | Problem-agnostic GA engine with pluggable encodings (binary, real, permutation, mixed), applied to classic OR problems. |
| ant-colony-optimization | ACO for discrete and continuous optimization on a shared metaheuristic core. |
All 222 Harzer Wandernadel hiking stamps, packed into about 70 one-day loop hikes of 10 to 20 km by harz-wanderung: shortest-path distances on the OpenStreetMap footpath network, ascent from the Copernicus DEM, constraint-aware clustering, and an exact closed-loop TSP from the nearest parking spot for every trip.
LinkedIn · Germany · Open to roles in supply-chain planning and optimization