Coinbase's ML Training Evolution: From Sagemaker to Ray - Ray Summit 2024

Explore Coinbase's transformative journey in ML infrastructure as they transitioned from Sagemaker to Ray in this 35-minute conference talk from Ray Summit 2024. Discover how this strategic shift dramatically accelerated ML training processes, reducing iteration times from hours to seconds and enabling scalability to terabyte-level datasets. Follow ML Platform Engineer Wenyue Liu's detailed account of the challenges faced with Sagemaker and the subsequent migration to Ray. Gain insights into how this transition not only enhanced training speeds and scale but also achieved over 20% cost reduction in instance usage. Acquire valuable knowledge for ML practitioners and engineers looking to optimize their training infrastructure, providing a blueprint for achieving unparalleled efficiency and scale in ML operations.

Coinbase's ML Training Evolution: From Ray to Sagemaker | Ray Summit 2024