# Purdue Analysis Facility

The Purdue Analysis Facility (Purdue AF) is an interactive environment for fast and scalable CMS physics analyses on dedicated computing resources at Purdue. You work in a personal JupyterLab or VS Code session, with CMS data, a maintained HEP software stack, GPUs, and Dask and Ray clusters within reach.

[Log in to Purdue Analysis Facility](<https://cms.geddes.rcac.purdue.edu/hub>)

Log in with a Purdue, CERN, or FNAL account — see [Login methods](<https://analysis-facility.physics.purdue.edu/login-methods/index.md>).

- **[Get started](<https://analysis-facility.physics.purdue.edu/getting-started/index.md>)**

  Create your first session and set up GitHub and Grid access, in about 15 minutes.
- **[Software](<https://analysis-facility.physics.purdue.edu/software/index.md>)**

  Use the shared Pixi environment with `coffea`, `ROOT` and `pytorch`, or build your own.
- **[Storage and data](<https://analysis-facility.physics.purdue.edu/storage/index.md>)**

  Know which volume to use for what, and read CMS datasets over XRootD, XCache and Rucio.
- **[Scale out](<https://analysis-facility.physics.purdue.edu/scaling-out/index.md>)**

  Run an analysis on hundreds of cores with Dask Gateway, or submit Slurm and CRAB jobs.
- **[Train on GPUs](<https://analysis-facility.physics.purdue.edu/gpus/index.md>)**

  Start a session with an A100 or T4 GPU, or train on several GPUs with Ray from a notebook.
- **[AI agents](<https://analysis-facility.physics.purdue.edu/agents/index.md>)**

  Claude Code, Codex and OpenCode run in every session and already know the facility.

Something not working? See [Troubleshooting](<https://analysis-facility.physics.purdue.edu/troubleshooting/index.md>) or [ask for help](<https://analysis-facility.physics.purdue.edu/support/index.md>).

![Purdue AF User Statistics](<https://cms.geddes.rcac.purdue.edu/users-graph/purdue-af-registered-users.png>)
