Overview

Multi-bank group wants shared risk analytics but can't share raw data (privacy, competition). Federated learning: models train locally on each bank's data, share gradients (aggregated). Central model improves without centralized data. Privacy preserved.