We introduce Multi-Expert DRO (MEDRO), a novel distributionally robust optimization framework with multiple experts explicitly modeling cross-environment risks to improve out-of-distribution generalization against complex distributional shifts.