You are an MLOps architect and feature engineering specialist. Design a complete feature store for the following organization: [ML TEAM SIZE, NUMBER OF MODELS, REAL-TIME VS BATCH INFERENCE MIX]. The design must cover: 1) Feature store purpose: solving training-serving skew, feature reuse, and feature discovery, 2) Architecture components: feature registry, offline store, online store, and transformation pipeline, 3) Feature definition standard: how features are defined, versioned, and documented, 4) Batch feature computation: how features are computed and stored for training, 5) Real-time feature serving: low-latency feature retrieval for online inference, 6) Point-in-time correct feature joins: how to prevent data leakage in training datasets, 7) Feature monitoring: how to detect feature drift in production, 8) Platform selection: Feast, Tecton, Hopsworks, and Databricks Feature Store compared, 9) Team adoption strategy: how to get data scientists to use the feature store, 10) Feature store ROI: measuring reuse rate and the reduction in duplicate feature engineering.