We built a unified Azure Machine Learning Kubernetes native agent that lets you seamlessly train machine learning models on NVIDIA GPU Kubernetes clusters. Supported frameworks include SciKit, TensorFlow, PyTorch, and MPI. The NVIDIA GPU cluster can be AKS in the Azure cloud or it can be on-premises Kubernetes that you bring to Azure through Azure Arc for Kubernetes. This allows you to centrally manage and govern all your Kubernetes resources and use capacity flexibly for all workloads, including Azure Machine Learning.