Traditional ML Ops platforms are not enough to develop models for autonomous driving. The challenge is not just about building a singular model efficiently, but how to develop an integrated ML stack capable of autonomous driving. In this talk, Kevin Haas, Head of ML infrastructure at Waymo, will discuss Waymo’s progress towards building a scalable technology platform for autonomous driving and give an overview of the ML operations and platforms behind it. Waymo has been operating the world’s first fully autonomous commercial ride-hailing service Waymo One in Phoenix, AZ since 2020. Now it is ferrying passengers with no human driver behind the wheel in multiple U.S. cities. Building powerful ML-based driving software and robust training and evaluation infrastructure has been key to scaling the technology across multiple geographies.