This document discusses machine learning workflows using Kubernetes and KubeFlow. It begins by explaining why Kubernetes is useful for machine learning for its composability, portability, and scalability. It then discusses using Jupyter notebooks for machine learning development workflows. It provides an overview of running machine learning on Kubernetes both with and without KubeFlow. Key components of KubeFlow like Kubeflow Pipelines are explained. Advantages of using KubeFlow on AWS are highlighted. TensorFlow and how AWS optimizes it is covered. The document concludes by discussing characteristics of autonomous vehicle workloads and how Kubeflow on AWS can help.