Scaling AI the Snowflake Way: ML Workloads on Ray | Ray Summit 2025

At Ray Summit 2025, Vinay Sridhar from Snowflake shares how Snowflake has integrated Ray as the backbone of its AI and ML stack—scaling to support thousands of models trained across terabytes of data within the Snowflake AI Data Cloud.

He begins by highlighting an industry-wide trend: Ray has become the de facto distributed computing platform for large-scale ML, DL, and RL workloads. Snowflake has embraced this shift, adopting Ray to power end-to-end training pipelines that operate seamlessly inside its unified data environment.

Vinay then walks through how Snowflake made this deep integration possible, focusing on the architectural decisions and platform enhancements required to bring Ray into the Snowflake ecosystem. A key part of this work involved enabling users to pull data directly from Snowflake tables, Apache Iceberg™, and other cloud storage sources into Ray with minimal friction—removing data silo boundaries and simplifying the path from data to model training.

Attendees will gain insight into how Snowflake uses Ray to scale machine learning across massive datasets, how Ray fits naturally into the AI Data Cloud vision, and how Snowflake customers can leverage this integration to build performant, scalable, and unified ML workflows.

Liked this video? Check out other Ray Summit breakout session recordings    • Ray Summit 2025 - Breakout Sessions  

Subscribe to our YouTube channel to stay up-to-date on the future of AI!    / anyscale  

🔗 Connect with us:
LinkedIn:   / joinanyscale