Powering Enterprise Feature Stores with a Universal Semantic Layer
summary
Claim: a universal semantic layer (AtScale example) lets enterprises populate and enrich feature stores so models actually deliver value. Demonstrated: a single-line Python connector to discover business-vetted measures (fiscal year, hierarchies), train with scikit-learn, then write back into semantic layer and publish to a feature store. Cited pain: only 1 in 2 ML models reaches production (Gartner) while spend nears $500B (IDC). Outcomes: consistent definitions, governed access, faster query performance, and reusable features for forecasting at a large retailer.
AI-written summary · about
chapters (7)
- 0:00Introduction to universal semantic layer
- 3:14Enterprise AI investment challenges
- 8:14Rethinking features with business metrics
- 11:25Semantic layer definition and benefits
- 14:18Integrating semantic layer with feature stores
- 18:36Practical example with forecasting model
- 21:05Conclusion and Q&A
related talks
30mHow Feature Stores Enable Operational ML
11mStrengthening the data value chain
32mDataEngBytes 2023 - PER-T1-11 - Mark Kitson
45mSemantic Layer: Make Your Data Speak with AI Without Hallucinations
27mEnterprise AI at Scale: Success Stories and Insights
30mDataEngBytes 2023 - MEL-T2-01 - Mikiko Bazeley
15mFeature Stores: Powering Machine Learning Infrastructure with Apache Flink
40mFrom Legacy to Building Our Own Feature Store: Active Steps