Applied Intuition's Blueprint for Scalable RL and Batch Inference

Learn how Applied Intuition scales massive inference and reinforcement learning workloads on petabytes of autonomous driving sensor data in this 31-minute conference talk from Ray Summit 2025. Discover Ray's role within Applied's ML infrastructure and how it enables unified, distributed execution across Kubernetes clusters. Explore how Ray Data powers large-scale batch inference pipelines by streaming raw sensor data from data lakes, executing CPU-intensive transformations, and seamlessly feeding results into GPU inference at scale. Understand how Ray's distributed execution model and RLlib support scalable open- and closed-loop reinforcement learning, including running thousands of parallel rollouts, colocating GPU learners with simulators for maximum efficiency, and restoring full training state with minimal overhead. Gain practical guidance from Applied Intuition's real-world experience managing Ray in production, providing valuable insights for teams applying Ray to large-scale inference and RL workloads in autonomous driving applications.