Evolution and Scaling of Feature Store at Uber

Evolution and Scaling of Feature Store at Uber

Explore the evolution and scaling of Uber's Palette Feature Store in this 33-minute conference talk by Divya Nagar from Uber Technologies. Learn how Uber introduced the concept of sharing features between teams and use cases to reduce storage and serving costs while maintaining high scalability and reliability standards. Discover how Palette currently serves over 21,000 features with 25 million queries per second at 99.999% uptime, and gain insights into the innovative approaches used to optimize feature management for machine learning models at scale.