# skypilot-org/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.

Repository: https://github.com/skypilot-org/skypilot
Canonical: https://ross.abutalabs.com/products/skypilot
Homepage: https://skypilot.ai/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: cloud-computing, deep-learning, gpu, hyperparameter-tuning, machine-learning, tpu, job-queue, job-scheduler, cloud-management, distributed-training, ml-infrastructure, multicloud, spot-instances, ml-platform, cost-optimization, llm-serving, llm-training, slurm, mlops
Last push: 2026-08-27T00:02:57+00:00

## Health v2 (maintenance only)
Score: 97/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 94, longevity 100
- inputs: {"age_days": 1848, "days_push": 7, "days_rel": 40, "gap_med": 18.0, "n_releases_24m": 25}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10529, forks 1211 (observed 2026-08-28T04:10:42.237001+00:00)

## What it is
SkyPilot is an open-source AI compute platform that unifies fragmented infrastructure (Kubernetes, Slurm, VMs, 20+ clouds) into a single pool for running, managing, and scaling AI workloads. It provides job scheduling, distributed training, LLM serving, and cost optimization across heterogeneous compute.

## Use cases
- run distributed training jobs across multiple clouds
- serve LLMs on my own GPUs
- find cheapest GPU cloud and launch there
- schedule ML jobs on spot instances
- manage a unified control plane over Kubernetes and Slurm clusters
- scale reinforcement learning inference on GPUs
- run batch inference jobs on any cloud

## When to choose
- you need to run AI/ML workloads across multiple clouds, Kubernetes, or Slurm without vendor lock-in
- you want gang scheduling, multi-node jobs, and spot-instance cost savings out of the box
- you need to train or serve LLMs on heterogeneous GPU fleets

## When to avoid
- you only deploy simple stateless web services on a single cloud
- you need a fully managed ML platform with no infrastructure to operate
- your workloads are non-AI batch jobs with no GPU or accelerator requirements

## Facets
- artifact type: framework
- maturity: active
- function: llm-training, llm-inference, scheduling, deployment, workflow-automation, gpu-computing, container-orchestration
- domain: machine-learning, deep-learning, cloud-computing, infrastructure-as-code, gpu-computing, large-language-models
- platform: python, cloud, cli
- tags: multicloud, job-scheduler, spot-instances, distributed-training, mlops, slurm, cost-optimization, ai-infrastructure, devops, kubernetes, linux, macos

## Member repositories
- skypilot-org/skypilot (main) score 97

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:42.237001+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:18:34.219673+00:00, confidence not recorded.
  - readme: https://github.com/skypilot-org/skypilot (fetched 2026-08-28T04:10:42.237001+00:00, sha ebb35341ef0a)
  - homepage: https://skypilot.ai/ (fetched 2026-08-29T08:17:53.850864+00:00, sha 1663aa87222c)
  - site_page: https://skypilot.ai/about (fetched 2026-08-29T08:17:53.860114+00:00, sha 98bea7a18b1c)
  - registry_pypi: https://pypi.org/pypi/skypilot/json (fetched 2026-08-29T08:17:53.862068+00:00, sha 37dd6408d827)
- Data as of 2026-08-30T08:39:29.467469+00:00.