AI / ML Systems
AI / ML Systems
Recommendation System Design
Two-stage retrieval + ranking, ANN candidate generation, two-tower embeddings, gradient-boosted ranker, A/B testing, and cold start strategies
Feature Store DesignOnline vs offline store, point-in-time correctness, training-serving skew, feature freshness trade-offs, and Feast / Tecton / Vertex AI architecture
ML Platform DesignTraining pipeline, model registry with lineage, online / batch / streaming serving, drift monitoring, and shadow / canary rollout