Fueling Innovation at Scale - Inside Pinterest's Machine Learning Platform

Discover how Pinterest's Machine Learning Platform drives innovation at scale in this 20-minute conference talk that reveals the architecture and strategies behind powering personalized experiences for millions of users worldwide. Gain insights into Pinterest's high-level approach to building ML platforms and explore the vibrant ecosystem that enables rapid experimentation and iteration of machine learning innovations. Learn about the comprehensive ML applications and platform overview, including the training and inference infrastructure that supports Pinterest's massive scale operations. Understand the ML iteration funnel approach that balances model versus data considerations, and explore how Pinterest standardizes model development through their MLM framework. Examine Pinterest's unified ML platform architecture and discover how training observability leads to significant efficiency gains across their systems. Explore optimization strategies for data iteration using Ray, and learn about new developer workflows that improve sampling efficiency. See how Weights & Biases is seamlessly integrated into Pinterest's ML lifecycle to support experiment tracking, model registry, and collaborative workflows, streamlining processes and empowering teams to deliver state-of-the-art ML solutions at Pinterest's demanding pace and scale.