AI workloads kept breaking whenever the GPU stack or cluster changed. This talk shows how the platform team designed versioned GPU environments on Kubernetes so the same code reliably produces the same result across time and teams.

Learn how platform teams build a hybrid AI platform on Amazon EKS using Hybrid Nodes, Terraform, and NVIDIA GPUs, providing self-service, opinionated defaults, and reliable GPU access for ML teams across cloud and on-prem.