Welcome to exalsius
exalsius is a platform for building and managing GPU infrastructure for AI workloads. It lets you assemble clusters from any mix of cloud providers and on-premise machines, then run development environments, inference endpoints, and distributed training jobs on top of them.
There are two ways to start, depending on which layer you want to operate:
Use the platform: CLI quickstart
Install the exls CLI, import a GPU node, deploy a cluster, and start a workspace in minutes. exalsius runs the management layer for you.
Run the operator yourself: operator quickstart
Turn two SSH-reachable Linux machines into a management cluster, a provisioned child cluster, and a running Jupyter workspace with the open-source operator and kubectl.
What you can do with exalsius
- Bring your own infrastructure: Add cloud instances or bare-metal machines to a shared node pool with node management.
- Run Kubernetes anywhere: Deploy multi-cloud and hybrid clusters for AI development and serving with cluster deployments.
- Launch AI workloads fast: Start notebooks, dev pods, inference endpoints, and training jobs with workspaces.
- Observe everything in one place: View metrics, logs, and traces with built-in observability dashboards.
- Operate it yourself: Describe clusters and workspaces as Kubernetes resources with the open-source exalsius operator.
How it works
exalsius follows a simple three-step workflow:
- Nodes: Connect the infrastructure you want to use.
- Clusters: Provision Kubernetes on selected nodes.
- Workspaces: Launch the environments and services your team needs.
Each step builds on the previous one. The CLI quickstart walks you through the full flow.
Underneath, clusters and workspaces are Kubernetes resources managed by the exalsius operator. The CLI and the platform create them for you; with the operator you can create them directly.
Understand the Colony, WorkspaceClass, and WorkspaceDeployment resources behind the platform.