Modernizing DoorDash Model Serving Platform with Ray Serve

Explore how DoorDash modernized its model serving platform using Ray Serve in this 31-minute conference talk. Learn about the ML Platform team's journey from a highly performant prediction service with limited framework support to a flexible, self-service ecosystem. Discover the reasons behind adopting Ray for model training and inference, and how Ray Serve's data scientist-friendly approach aligned with DoorDash's core ethos. Gain insights into the evaluation process of various frameworks and the transition from earlier generation prediction services to the Ray Serve ecosystem. Understand the challenges faced in supporting a wider range of libraries for NLP and image processing use cases, and how the new platform addresses these needs. Access the accompanying slide deck for visual aids and additional information on this transformation in ML model serving at DoorDash.