SkyPilot

SkyPilot enables the execution of distributed training, batch processing, and inference tasks across diverse environments. It abstracts infrastructure complexity, allowing users to run compute-intensive jobs on Kubernetes, Slurm, or public cloud providers without manual configuration of underlying nodes or networking.

SkyPilot is designed for machine learning engineers, infrastructure architects, and data scientists who manage high-performance compute resources. It is ideal for teams needing to scale research or production workloads across multiple cloud regions or hybrid environments while maintaining consistent job definitions.

To use SkyPilot, you must have access to at least one supported cloud provider account, such as AWS, GCP, or Azure, or an existing Kubernetes or Slurm cluster. The environment requires a local installation of the package to interface with your cloud credentials and manage job submission.