Users (tenants) to express intent through model serving (20+ models), jobs (serverless), or GPU instance creation requests with automated instance management to scale, recover, or auto-stop.
GPUaaS and private cloud providers can dynamically buy GPUs from partner GPUaaS providers to never run out of capacity. Similarly, sell underutilized capacity to to increase revenue.
Create, observe, and manage bare-metal, virtual machine, or Kubernetes clusters with network, Infiniband, storage, compute, GPU isolation.
Hardware onboarding, tenant management, and image/catalog management with support for NVIDIA NCP and storage reference architectures. Advanced features include disk wiping, automated system tuning, fault management for reduced MTTR, cloud adjacent storage, and more.
.png)
