Running AI/ML workloads on Kubernetes shouldn't be a black box. This session is a practical, live-fire demo of scheduling a machine learning workload on a GPU-enabled Kubernetes cluster. We will start from scratch, configure a node pool with GPUs, and walk through the essential manifest configurations—including taints, tolerations, and resource limits—needed to make it work. As a Google Cloud AI Infra specialist, I'll highlight common "gotchas" and share best practices learned from helping customers run AI at scale. You'll walk away with the practical knowledge to start deploying your own AI applications on Kubernetes.