Feature Store Usage Patterns - From a Single Data Scientist to an Enterprise
summary
Claims that a feature store mainly solves the organizational life cycle not heavy lifting compute and demonstrates a virtual feature store that layers on top of heterogeneous infra. It runs in local mode with laptop compute and file system storage as a gitlike repo for feature and training set definitions. It separates processing needs (10 ms latency, 100 TB, Redis, Spark, Snowflake) from governance needs and provides immutable definitions, access control, lineage, and legal signoff workflows used at enterprise scale.
AI-written summary · about
chapters (5)
- 0:00Introduction and background
- 1:00ML platform vs MLOps
- 2:15The virtual feature store
- 4:15Benefits of feature stores
- 5:49Enterprise scaling and wrap-up
related talks
30mHow Feature Stores Enable Operational ML
20mNavigating the Skies of Data: A Flight Through Feature Stores
32mA feature Store for Machine Learning
33mFeathr, an Open Source, Battletested, Scalable and Enterprise-Grade Feature Store
15mFeature Stores: Powering Machine Learning Infrastructure with Apache Flink
19mML feature store a cornerstone in modern data architecture for AI/ML
30mDataEngBytes 2023 - MEL-T2-01 - Mikiko Bazeley
32mBuilding a Cloud Native Feature Store with Feast on Kubeflow