Ved Mohan

Research

A Fair OR–ML Framework for Resource Substitution in Large-Scale Networks

Ved Mohan, El Mehdi Er Raqabi, Pascal Van Hentenryck · Under revision at Transportation Science

Swapping equipment on already-scheduled trips to fix freight imbalances, fairly across zones. ML cuts model size by 80% and solve time by 90%, with the same optimum.

Map of the lower 48 with about 1,700 freight locations, the lanes between them, and 46 colored zones.
About 1,700 locations, 46 zones.

Democratizing Large-Scale Re-Optimization with LLM-Guided Model Patches

Tinghan Ye, Arnaud Deza, Ved Mohan, El Mehdi Er Raqabi, Pascal Van Hentenryck · NeurIPS 2026 Workshop on ML×OR

Plain-language requests become auditable model patches, then a re-solve. Tested on supply-chain re-optimization and exam scheduling.

Diagram of the ReOpt-LLM framework: an LLM sits between the user interface, the optimization model, a toolbox of re-optimization techniques, and the optimizer.
Request → model patch → re-solve.