Kim Hazelwood, Head of AI Infrastructure, Facebook Machine learning sits at the core of many essential products and services at Facebook. This presentation will describe the hardware and software infrastructure that supports machine learning at global scale. Facebook’s machine learning workloads are extremely diverse: services require many different types of models in practice. This diversity has implications at all layers in the system stack. In addition, the amount of data flowing through machine learning pipelines presents practical challenges in delivering data to high-performance distributed training flows. Computational requirements are also intense, leveraging abundant general-purpose and more limited specialized computing platforms for training and real-time inference. Addressing these and other emerging challenges continues to require diverse efforts that span machine learning algorithms, software, and hardware design.