Kick-off
summary
Hopsworks claims to be the first open-source, enterprise, and serverless feature store and cites Uber Michelangelo as the origin of the idea. The feature store bridges SQL warehouses and Python ML by separating production into three pipelines: feature (including backfill), training, and inference, supports real-time event data that bypasses the data warehouse, and stores features plus predictions to enable a feedback loop for data-centric AI. It also positions feature stores as the short-term memory or key-value associative memory for autonomous systems.
AI-written summary · about
chapters (5)
related talks
32mA feature Store for Machine Learning
30mHow Feature Stores Enable Operational ML
33mFeathr, an Open Source, Battletested, Scalable and Enterprise-Grade Feature Store
20mNavigating the Skies of Data: A Flight Through Feature Stores
33mEvolution and Scaling of Feature Store at Uber
34mSolving the ML Pain You Forgot to Mention
40mCreating and operating ML models from event based data using feature stores
15mFeature Stores: Powering Machine Learning Infrastructure with Apache Flink