skypilot

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faste

What is skypilot?

The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.

Overview of skypilot

What it is

SkyPilot is a system for running, managing, and scaling AI workloads on any AI infrastructure. It is designed for AI teams to use a simple interface to run jobs on any infrastructure, while infrastructure teams get a unified control plane for managing AI compute with advanced scheduling, scaling, and orchestration.

Key features

  • Easy to use for AI users: quickly spin up compute, environment and job as code, easy job management with queuing, running, and auto-recovery.
  • Makes Kubernetes easy for AI and infra teams: Slurm-like ease of use, cloud-native robustness, local dev experience on K8s with SSH, code sync, or IDE connection, and advanced scheduling like gang scheduling, multi-cluster, and scaling.
  • Unifies multiple clusters, clouds, and hardware: one interface for reserved GPUs, Kubernetes clusters, Slurm clusters, or 20+ clouds, flexible provisioning of GPUs, TPUs, CPUs with smart failover, and team deployment and resource sharing.
  • Maximizes GPU fleet utilization: autostop for idle resources, binpacking on shared clusters, and intelligent scheduler that automatically schedules on the most available infrastructure.
  • Supports existing GPU, TPU, and CPU workloads with no code changes.

Use cases

SkyPilot can be used for training, serving, running LLM models, AI applications, and common frameworks. Example use cases include training with Verl, Finetune Llama 4, TorchTitan, PyTorch, DeepSpeed, NeMo, Ray, Unsloth, Jax/TPU, OpenRLHF; serving with vLLM, SGLang, Ollama; running models like DeepSeek-R1, Llama 4, Llama 3, CodeLlama, Qwen, Kimi-K2, Kimi-K2-Thinking, Mixtral; AI apps such as RAG and vector databases; and common frameworks like Airflow, Jupyter, and marimo.

Getting started

Install with uv (also supported: pip, nightly, from source):bashuv pip install 'skypilot[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,seeweb,shadeform,verda]'Launch your first cluster by creating a task file my_task.yaml and using sky launch my_task.yaml. See the repository README for the full quickstart guide.

License and pricing

SkyPilot is open source under the Apache-2.0 license.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/skypilot-orgskypilot.svg)](https://yunodb.com/tools/skypilot-orgskypilot?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/skypilot-orgskypilot?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/skypilot-orgskypilot.svg" alt="Featured on YunoDB" height="20"></a>