Combinatorial Optimization for Robust AI Resource Allocation and Safe Model Deployment

ScopeProblem scope: offline, deterministic resource allocation for AI infrastructure across three tasks: (i) data-storage assignment, (ii) pipeline routing with capacities, (iii) conflict-free resource allocation.Models: bipartite matching (assignment), max-flow (routing), and general matching (compatibility graph).Evaluation: compare against greedy/single-path baselines across problem sizes (10-500) using latency, throughput, feasible pairings, and conflict counts.Out of scope: online/adaptive scheduling, stochastic demand, distributed consistency, multi-objective fairness, and approximate/learned optimizers.