The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) ML Feature Store at Intuit with Srivathsan Canchi - #438
Dec 16, 2020
Srivathsan Canchi, Head of Engineering for Intuit's Machine Learning Platform, discusses the groundbreaking role Intuit played in developing the AWS SageMaker Feature Store. He explains how feature stores enhance machine learning by ensuring data consistency and addressing challenges like feature drift. The conversation also touches on the exploding interest in feature stores, their importance in scalable AI, and the implementation challenges faced during their establishment, including the benefits of GraphQL integration. Tune in for insights on the future of machine learning!
AI Snips
Chapters
Transcript
Episode notes
Cash Flow Prediction
- Intuit uses ML to predict cash flow for small businesses, especially helpful during COVID's uncertainties.
- This empowers customers to make informed financial decisions.
Genesis of Intuit's Feature Store
- Intuit's various business lines (taxes, accounting, personal finance) intersect and share customer data.
- This drove the need for a feature store to enable efficient feature sharing across models.
Importance of Feature Consistency
- Feature discovery and sharing enable training on relevant features.
- Consistent feature usage across training, evaluation, and inference is also crucial for model reliability.