Scaling User-Focused Foundation Models at Grab with Ray

Explore how Southeast Asia's leading super app builds powerful user embedding foundation models to understand complex, multi-modal user journeys across diverse services. Learn from Nick Buehrer and Chongyu Zhou from Grab as they demonstrate their approach to developing unified models that capture holistic user behavior across ride-hailing, food delivery, payments, and other interconnected services. Discover how Grab creates general-purpose user embeddings that serve as universal numerical representations, eliminating the need for siloed, hand-engineered features while enabling more personalized, consistent, and scalable machine learning capabilities across their platform. Gain insights into multimodal modeling techniques, foundation models for personalization, and large-scale production ML systems through this detailed examination of Grab's innovative approach to understanding the rich behavioral footprints of users in a complex super app ecosystem.