Start a Pilot

A pilot begins with your Slurm scheduler export and four questions about your facility. We run the analysis and deliver a private assessment report to your team. The data collection takes under two hours for a standard Slurm export.

What a pilot covers

A standard pilot runs ACE, PACE, and GRADE — the core efficiency engines that require only a Slurm scheduler export. If you can share facility information (cooling type, power source), we add COOL and FLUX. If you have hardware inventory data, we add CORE. You receive a PTL Score and full ATLAS recommendations for every engine we run.

What you provide

• A Slurm sacct export covering at least 30 days of job history
• Facility basics: cooling type, location, grid carbon intensity estimate
• Optional: hardware inventory, power source documentation

If your Slurm export contains sensitive job names or user identifiers, we work with you to anonymize before transfer. CLAW’s local-only mode computes metrics inside your environment — only the computed numbers transit to PTL.

What we deliver

Within five business days of receiving your data:

• Full HTML assessment report — PTL Score, tier, engine findings, ATLAS action plan
• Structured JSON exports for all engines you provide data for
• A working session to review findings together

You decide what to do with the results. We do not publish pilot findings without explicit permission. Follow-up assessments can use the pilot findings as the baseline.

Start a pilot

Write to research@plaintheory.org. Include a brief description of your cluster — scheduler type, approximate GPU count, and primary workload category. We will send the pilot intake form within one business day.