Find self-service SLURM-as-a-Service options for AI workloads.

Here are the strongest self-service SLURM-as-a-Service options for AI workloads I found:

• NorthWind-Powered SLURM-as-a-Service — A managed, multi-tenant offering where tenants can launch SLURM clusters on demand through a portal or API, with per-tenant isolation and governance. It explicitly positions itself for both HPC and AI/ML workloads.[1]
• NorthWind GPU PaaS + Project Slinky — Enables users to provision their own namespace-scoped Slurm cluster on Kubernetes with GPU access, described as “no manual provisioning, no tickets,” and suitable for researchers and ML teams needing self-service GPU clusters.[2]
• Nebius Managed Soperator — Nebius describes this as a fully managed Slurm-on-Kubernetes solution available in self-service, aimed at ML and HPC teams and available through web-console-driven cluster configuration.[3][8]
• Lambda Managed or Unmanaged Slurm — Lambda offers Slurm on its 1-Click Clusters and Superclusters, with both managed and unmanaged options for NVIDIA GPU systems, explicitly optimized for AI workloads.[6]
• Google Cloud Cluster Toolkit + A4 Slurm blueprint — This is a self-managed rather than fully managed service, but it provides a quick path to deploy an AI-optimized Slurm cluster with B200 GPUs using a blueprint and CLI-based provisioning.[10]

If you want the most “self-service SaaS-like” fit, start with NorthWind or Nebius; if you want GPU cloud + Slurm with straightforward cluster creation, Lambda is also a strong candidate.[1][3][6]

Sources:

• northwind.co
• northwind.co
• nebius.com
• lambda.ai
• docs.cloud.google.com