AID2E General Meeting

Date: Aug 11, 2026

Announcements

• XLOOP 2026 artifact submission deadline was July 31. The team will look for other conference and workshop opportunities.
• MODE 2026: September 1, 2026.

Updates on Phase I from technical meeting

Derek — JLab

• PR #57 (“Resolve config issues”) was merged on July 29.
• PR #22 was merged, so new PRs should now target main.

• PR #59 needs to be updated to target main.

• The aid2e optimize command now handles:

  • Optimization loop
  • Scheduler execution
  • Objective collection
  • Failure reporting
  • Result artifacts
• Users provide a YAML configuration file, an analysis script, and optional helper code.
• Next two weeks (around August 25):
  • Make sure the optimization pipelines work, including dRICH on SLURM and PanDA.

Amit — BNL

• PanDA and iDDS scheduler implementations are in progress.
• Simple PanDA jobs have been successfully implemented and tested.
• Multi-stage PanDA jobs still need to be finalized.

Code Assist

James — W&M

• Continued work on the Code Assist branch.
• Once the implementation and examples are complete, it will be ready for testing.
• Testing is waiting on Baptiste's DUO authentication issue.
• Baptiste reported that the issue is on the W&M side, and the team is switching from DUO to Microsoft Authenticator.

Monitoring / Control

Amit — BNL

• Testing of inference using the vLLM server on Perlmutter will resume when the system is available.
• Tianle has a multi-model vLLM server that can be deployed on Perlmutter.
• Amit will meet with Tianle on Thursday to discuss integration and testing with the Monitoring agent.
• The setup is mainly intended for Perlmutter jobs, though some parts may be reusable by other institutions.

Monitoring Goals

• Use AID2E and PanDA logs and outputs to monitor optimization jobs.
• Identify why jobs fail and at which stage.
• If a failure is likely to affect other jobs, suggest stopping queued jobs.
• Use previous optimization results to suggest new design points.
• Allow offline searches of past optimization jobs, including failures, results, and performance information.
• Add PanDA support for accessing locally stored optimization results.
• Temporarily store PanDA logs centrally for access to ongoing and recent jobs.

Detector Optimization Applications

CUA

• Calorimetry calibration: Updates requested from Baptiste and Tanja.

Duke / W&M / Other Institutions

• kLM Design: Phase I activity.
• RICH Alignment: Phase II activity; alignment is in production and included in Connor's thesis.
• AerFib: Ongoing activity.
• Software Stack Extension: FEM integration activity.