Microsoft’s Azure Kubernetes Service engineering team has open-sourced TauGrid, an MIT-licensed, self-hosted platform for running AI workloads on Kubernetes, published with container images and Helm charts as public OCI artifacts on Microsoft Container Registry. TauGrid combines five usually hand-integrated components: the tau CLI, workload queueing and admission through Kueue, Ray cluster orchestration through KubeRay, node-level GPU health monitoring, and cluster and workload observability. It requires a Kubernetes 1.30+ cluster with GPU nodes, kubectl, and Helm 3.0 or later, and installs via a Helm chart from MCR at version 0.4.2 into the tau-system namespace. Workloads are described in a tau.yaml file; on tau run, TauGrid resolves platform policy, renders a Kubernetes Job or KubeRay RayJob, and submits it through Kueue. Microsoft documents six stages: submission, queueing, execution, monitoring, recovery, and evidence. Recovery covers retry, resume from checkpoint, and failure diagnosis, while evidence records capture workload metadata, configuration, logs, metrics, checkpoints, and execution history. Platform teams own workspaces, queues, compute profiles, storage, identity, and observability; researchers submit workloads from a repository and CLI without configuring Kubernetes directly. The codebase is primarily Go. First-party images ship under mcr.microsoft.com/aks/ai-runtime/ for Tau, the TauGrid Portal, and