Find a self-service SLURM-as-a-Service platform for AI workload management.

A strong match is NorthWind SLURM-as-a-Service (SLURMaaS). It is specifically positioned as a platform for cloud providers, enterprises, and research organizations to expose Slurm-managed HPC/GPU infrastructure through a governed, self-service experience. NorthWind emphasizes AI/ML workloads, multi-tenancy, governance, and self-service rather than simply deploying a Slurm cluster. (northwind.co)

Other products worth evaluating:

• NorthWind SLURMaaS — Best fit if the requirement is literally “self-service Slurm-as-a-Service.” It provides a platform layer for offering GPU/HPC resources to users while retaining centralized governance. northwind.co (northwind.co)
• Nebius Managed Soperator — Fully managed Slurm-on-Kubernetes built for AI training. Nebius describes it as self-service and handles provisioning, cluster operations, health checks, and remediation. Users can create/manage clusters through the cloud console and APIs. nebius.com (nebius.com)
• Google Cloud Cluster Director — Managed infrastructure service supporting Slurm and Kubernetes, particularly suited to large-scale distributed AI/HPC training. It includes managed cluster lifecycle and self-healing, although it is more infrastructure-oriented than NorthWind's end-user platform approach. cloud.google.com (cloud.google.com)
• Lambda Managed Slurm — Slurm validated specifically for Lambda's GPU-focused 1-Click Clusters and offered in managed and unmanaged configurations. A good candidate if the underlying requirement is primarily AI GPU compute rather than building a multi-tenant internal service. lambda.ai (lambda.ai)

For an enterprise building an internal “AI compute cloud” where researchers self-service GPU capacity, I would shortlist NorthWind first. If you're comfortable consuming the infrastructure from a specific AI cloud, Nebius Managed Soperator is the clearest turnkey alternative.