From bare metal to full-stack AI platform
A Complete AI Platform on Mirantis k0rdent Infrastructure
Mirantis k0rdent turns bare metal GPU servers into managed Kubernetes clusters. Saturn Cloud turns those clusters into an AI development platform. Together, you get the platform experience of a hyperscaler on infrastructure you control.
Why Mirantis + Saturn Cloud
GPU providers need platform layers,
not just compute
GPU providers compete on hardware, but customers expect the platform experience they get from hyperscalers – not SSH access and support tickets. k0rdent manages the infrastructure layer, Saturn Cloud provides the platform layer, and together you get real operational benefits: higher utilization, self-service revenue, and less support overhead.
Monetize spare capacity
Self-service access lets you sell on-demand GPU hours between reserved commitments. Instead of machines sitting idle between contracts, customers can launch workloads instantly through web UI or API – turning spare capacity into revenue.
Secure multi-tenancy
k0rdent creates dedicated Kubernetes clusters per tenant with isolated network fabric and dedicated nodes. When a customer churns, machines are wiped and returned to the pool.
GPUs operational on day one
k0rdent turns bare metal into managed Kubernetes clusters using Metal3 and Cluster API. Saturn Cloud installs via Helm charts in under 30 minutes. No manual imaging or VLAN configuration.
Visibility and idle detection
Track GPU utilization by user and project. Automatically shut down forgotten notebook servers. Identify idle resources instead of billing customers for GPUs sitting dark.
How it works
The combined architecture
k0rdent manages infrastructure
Orchestrates bare metal servers via BMC protocols. Creates Kubernetes clusters using Metal3 and Cluster API. Handles server imaging, network configuration, and cluster lifecycle across hundreds of nodes.
Saturn Cloud provides the platform
Installs into k0rdent-managed clusters as Helm charts. Delivers JupyterLab, VS Code, job scheduling, model deployments, distributed training coordination, SSO, and usage tracking.
You provide the hardware
GPUs, networking, and storage. k0rdent and Saturn Cloud handle everything else. Infrastructure engineers manage k0rdent; AI/ML engineers use Saturn Cloud. Neither team needs to know the other's tools.
The shift
From manual provisioning to automated platform
| Without k0rdent + Saturn Cloud | With k0rdent + Saturn Cloud |
|---|---|
| Email threads to provision servers | API calls to create resources |
| Manual server imaging and VLAN configuration | Automated via Metal3 and Cluster API |
| Send customers IP addresses and SSH keys | Send customers a login URL |
| Customer installs drivers, CUDA, networking stack | Pre-configured in cluster images |
| Customer configures InfiniBand/RoCE for multi-node | k0rdent provides RDMA drivers, Saturn Cloud orchestrates jobs |
| No visibility into GPU utilization | Usage tracking and idle detection built-in |
| Support tickets for "How do I run Jupyter?" | Self-service dev environments |