Skip to main content

Bring Your Own Compute

By default, calculations run on Atomiverse-managed infrastructure. Bring Your Own Compute (BYOC) lets you run jobs on your own hardware instead — a workstation, a local server, or an HPC cluster.

When BYOC Makes Sense​

  • You have existing compute capacity you want to use
  • Your data needs to stay on your own infrastructure
  • You run large jobs that benefit from your own hardware
  • You have access to an HPC cluster with SLURM

How It Works​

You install the atomiverse-controller on your machine and start it. The controller connects to Atomiverse, picks up any jobs you submit with your compute target specified, runs them locally using Docker or Apptainer, and sends results back.

From Python, the only change is one extra argument when submitting:

job.submit(compute_target="my-cluster")

Everything else — the same SDK, the same result format, the same job lifecycle — works identically. See the controller setup guide for the full list of submission options including capacity waiting and timeouts.

Getting Started​

→ Set up the controller → Estimate controller usage — inspect free capacity and predict how many jobs will fit