Cedana

Cedana is an automation engine for AI and HPC orchestration that runs workloads faster, with higher throughput, and more reliably by dynamically moving jobs across cloud and on-prem resources without losing any work. We do this by: * First, plugging into existing systems: Kubernetes, SLURM, Ray, Kueue, Kubeflow - no need to change your stack or code. * Then, creating a global compute pool: treats all your on-prem and cloud infrastructure as one, easy to use, global cluster. * Lastly, making all workloads SAFELY elastic and durable: never lose work. It moves running jobs (including GPU-intensive ones) between machines, clusters, or clouds with absolutely no work lost. Jobs automatically continue through HW failures, OS updates, spot recovations, etc. Resize, suspend, and resume workloads at any time. The result of this is: Teams get 2-10x improvements in throughput and productivity, mission-critical reliability, with dramatically faster time-to-insights. The benefit is: Cedana unlocks full infrastructure efficiency so teams can innovate faster, reduce waste, and spend more time building instead of firefighting.