Skypilot

SkyPilot is an AI compute platform that lets teams run and manage workloads across Kubernetes clusters, Slurm clusters, virtual machines, on premises hardware and more than 20 clouds from a single interface. It began as an open source project and now also offers SkyPilot Platform for organizations that need shared governance over GPU resources. The platform covers the working pieces of AI infrastructure: a command line interface for launching jobs, an intelligent scheduler, gang scheduling and multi node jobs, SSH access, GPU monitoring, and quotas and policies with a dashboard so that many researchers can share capacity without a ticket queue. The goal is to let a team move between clouds and cluster types without rewriting how jobs are launched and without being locked into one provider. It is aimed at machine learning engineers, AI researchers and infrastructure teams running training, reinforcement learning and other GPU heavy work at scale. The vendor publishes case studies from companies that moved from Slurm to a multi cloud setup. The open source version can be tried directly. The vendor does not publish prices for SkyPilot Platform, and access is arranged by booking a demo. The company announced a 20 million dollar seed round when it launched the platform.