Whenever you're ready, walk me through how you'd approach this.
How would you architect a feature store to be shared across multiple AI engineering teams at GEICO?
Discuss the offline and online serving layers, batch and streaming feature pipelines, point-in-time correctness, feature ownership, discoverability, access control, and training-serving consistency. Explain how you would support backfills, schema evolution, data quality, lineage, monitoring, and failure recovery without allowing one team’s workloads to compromise another team’s features.