Amazon SageMaker AI Spaces brings JupyterLab to EKS in 5 minutes

Amazon Web Services has introduced the SageMaker AI Spaces add-on for Amazon EKS, which runs interactive IDEs like JupyterLab and Code Editor directly on the same Kubernetes cluster hosting data pipelines. Previously, data scientists had to leave the cluster to use standalone JupyterHub deployments or local laptops, losing access to GPU nodes, shared storage, and IAM roles. The new add-on closes that gap by provisioning fully configured environments without needing a separate platform team effort—what typically took 3–5 days to stand up now takes about 5 minutes per Space.

The solution layers networking, routing, and compute on a single EKS cluster: Route 53 points a wildcard domain to an Application Load Balancer with TLS, Traefik handles hostname routing, and AWS KMS encrypts JWT tokens for auth. For VS Code users, AWS Systems Manager tunnels directly to Space pods, and EKS Pod Identity grants scoped IAM roles. Under the hood, private-subnet workers run Space pods with Amazon EBS for persistent volumes and EFS or FSx for shared storage. By consolidating interactive and training workloads, organizations can keep GPU nodes busy between jobs, lifting utilization by up to 30% and avoiding the thousands of dollars in monthly costs from always-on GPU notebook environments.