Berkelium: High-Performance Kubernetes for LBNL Science
The dedicated, multi-tenant Kubernetes platform engineered by LBNL Science IT to empower laboratory scientists with elastic container orchestration, 32 enterprise GPUs (NVIDIA H100, A100, A40, AMD Instinct MI100), Coder IDEs, and high-throughput scientific data pipelines.
Open Namespace Portal 🚀
Accelerated Scientific Workloads & GPUs
Berkelium connects 32 enterprise GPUs across NVIDIA (H100 80GB HBM3, A100 80GB PCIe, A40) and AMD (Instinct MI100 ROCm) directly to modern containerized workflows. Run distributed PyTorch, TensorFlow, AlphaFold, Ray, or MPI jobs with native Kubernetes scheduling.
- NVIDIA GPU Operator with Multi-Instance GPU support
- AMD ROCm GPU support for heterogeneous scientific code
- Interactive Coder workspaces & JupyterLab notebook environments
Self-Service Multi-Tenant Namespaces
Scientists and research groups can provision dedicated, isolated Kubernetes namespaces in seconds authenticated via your institutional LBNL OneID credentials.
- Instant self-service namespace provisioning
- LBNL LDAP / Single Sign-On (SSO) authentication & RBAC
- Automated kubeconfig and CLI token generation
Backed by LBNL Science IT Expertise
Our team of computational scientists and infrastructure engineers provides direct consultation, containerization support, and workflow optimization to help your project thrive.
- Hands-on container migration (Docker, Apptainer, Singularity)
- Weekly office hours
- Continuous monitoring via Prometheus/Grafana metrics
Ready to run your scientific computation on Berkelium?
Join hundreds of LBNL researchers scaling their science. Sign in with your LBNL credentials, claim your project namespace, and download your kubeconfig in minutes.
Enter Namespace Portal 🚀