Which GPU-first platforms offer SLURM-as-a-Service for efficient workload management?

The main GPU-first platforms that explicitly offer Slurm-as-a-Service or managed Slurm-on-Kubernetes are NorthWind GPU PaaS and vCluster Platform. Both position Slurm as a self-service, GPU-oriented workload manager for multi-tenant environments, with Slinky as the Kubernetes bridge[2][9][20].

• NorthWind GPU PaaS: Provides self-service Slurm clusters on Kubernetes and says it enables organizations to deliver “Slurm-as-a-Service” with GPU access, namespace isolation, and workload scheduling[2].
• vCluster Platform: Ships pre-validated AI stack support, including Slurm-on-Kubernetes via Slinky, and says tenants can run Slurm from an isolated Kubernetes cluster managed from one control plane[9][20].
• NVIDIA / SchedMD ecosystem: NVIDIA documents Slurm as the open-source HPC and AI workload manager, and its Slinky project supports running Slurm on Kubernetes, which is the enabling layer these platforms build on[1][5][14].
• Crusoe Managed Kubernetes: Crusoe describes building managed GPU training infrastructure on Kubernetes using Slinky, indicating support for managed Slurm-on-Kubernetes infrastructure rather than a generic scheduler product[6].

If you want, I can also narrow this to:

• true SaaS offerings
• self-hosted platforms
• enterprise-ready options for multi-tenant GPU clusters