Make with H2O.ai: Getting Started with H2O AI Feature Store
- Predictive AI
- H2O Feature Store
- H2O Driverless AI
Overview
In this session, H2O.ai's head of products and AT&T's Prince Paul Raj, AVP of Data Science in AT&T's Chief Data Office, walk through why AT&T partnered with H2O.ai to build the H2O AI Feature Store. They explain the problem of duplicated, inconsistent, and hard-to-audit features across data science teams, then demo how the feature store centralizes, versions, and shares features across offline and online use cases. The session closes with a live demo covering feature search, SQL and Power BI access for citizen data scientists, Driverless AI integration, and real-time scoring.
Key takeaways
- Why data science teams end up rebuilding the same features repeatedly, and how a feature store solves reuse and discovery
- How AT&T uses the feature store to track feature lineage and versioning for model audits and reproducibility
- The difference between the offline and online store, and how time travel lets you recreate point-in-time snapshots for backtesting
- How data engineers, data scientists, citizen data scientists, ModelOps, and compliance teams all work from the same feature store
- How to register a feature set, ingest data, and connect it to H2O Driverless AI to publish derived features back to the store
- How to expose features for real-time scoring, including a live demo of an online lookup and a feature-store-powered chat agent
Start owning your AI