Managed Kubernetes
Kubernetes at the Scale of the Fleet
GPU-native Kubernetes with a managed control plane. Radiant operates the control planes, node pools, and accelerator stack. You run your workloads.
Create, scale, upgrade, isolate, and recover control planes through automated lifecycle workflows. Upstream Kubernetes, no forks, defined EOL windows, patches applied automatically with minimal-downtime upgrades.
Accelerator and operator control
Manage GPU, network, CSI, DRA, and infrastructure operators, with tenant-specific driver and accelerator configurations.
Dynamic Resource Allocation
Categorize, request, and share GPU resources across pods.
Operators
Run standard accelerator operators; override default GPU or network drivers.
Storage
CSI for block, shared file, and NFS, with dynamic provisioning, snapshots, and resizing.
Node pool management
Create, scale, label, taint, and upgrade CPU and GPU node pools across tenant clusters. Instance-level control, auto-scaling to thousands of GPUs.
Networking and isolation
Standards-compliant CNI, Calico, multiple clusters per VPC, dual-stack IPv4/IPv6
Control-plane isolation, endpoint controls, IAM, encrypted state
Control-plane logs and Prometheus-ready API server metrics
