Learning to Divide-and-Conquer for Large Scale Vehicle Routing Problems

Scope

Build a learning‑guided divide‑and‑conquer pipeline for large‑scale vehicle routing, blending GNN/RL components with classical OR heuristics to deliver fast, high‑quality solutions on real logistics data.

What you’ll do

Prototype partitioning and search orchestration, integrate with established heuristics, and benchmark/optimize runtime and solution quality.

Profile

• Excellent Python
• Solid ML fundamentals
• Exposure to RL and/or GNNs
• Comfort with combinatorial optimization/MILP/heuristics
• C++ and parallel/distributed computing are a plus

Details

• Zurich‑based, mostly in‑person preferred.
• Start: As soon as possible
• Co-supervised with Prof. Daniel Kuhn and RAO Lab @ EPFL

Apply (EPFL only)

• Send CV, transcripts, and a motivation letter to [email protected], [email protected] and [email protected]
• More details upon request 🙂