AID2E General Meeting

Date: July 28, 2026

Announcements

Updates on Phase I from technical meeting

Cynthia — AID2E Framework

• Worked on the CLI; PR #57 has been approved.
• After PR #57 is merged, the remaining changes will be merged into main.
• Future PRs will build on the updated CLI implementation.
• The team is transitioning toward a common framework for all design optimization pipelines.
• The revised project timeline will be discussed at the next technical meeting.
• Completing the CLI is expected to be the final major step before moving forward.

Phase II

Code Assist

James — AID2E

• Created a new branch for the Code Assist work.
• Added deployment instructions under src/aid2e/mcp/.
• Testing is currently waiting on DUO authentication issues with Baptiste's William & Mary account.
• Testing will begin once the authentication issue is resolved.

Monitoring / Control

• Amit

  • Continued work on the PanDA jobs and agent framework.
  • The system currently supports two types of MCP tools:
  • MCP tools developed by the ePIC teams.
  • MCP tools developed specifically for AID2E.
  • Two approaches are being considered for the AI agent:
  • A custom AI agent orchestrator.
  • Existing tools such as Codex or Claude Desktop.
  • The custom AI orchestrator works with both:
  • Frontier models.
  • Open models hosted locally or through a vLLM server on Perlmutter.
  • Using Codex with the vLLM server on Perlmutter is still being worked on.
  • Further testing will resume once Perlmutter is back online in early August.

  • The custom AI orchestrator is currently working well; integration with Codex and similar tools remains in progress.

  • Continued work on deploying the MCP server locally.

  • Deployment is working on NPPS machines that can access the PanDA API through the secured connection.

• Wen

  • Working on saving optimizer results as JSON.
  • The JSON results will be accessible through the PanDA REST API and used by the PanDA MCP server together with the PanDA database.
  • Working on creating and indexing a database using OpenViking.
  • The goal is to allow the LLM to search optimization results and logs more efficiently.

Detector Optimization

Calorimetry Calibration

CUA

• Working on the calorimetry calibration study.
• The π0 sample shows a dependence on energy.
• Further investigation is ongoing.
• A technical meeting will be used to discuss the optimization strategy and how it can be supported within the AID2E framework.

Conference Papers

• Conference paper activities are ongoing.
• The XLOOP 2026 conference was mentioned for potential paper submission/work.