Kim Meagher

Advisor: Michael E. Papka

Mentor: Taylor Childers

Undergraduate: University of Michigan (Computer Science Engineering)

Graduate: University of Illinois Chicago (Computer Science)

Project: Agentic Workflow Optimization for Parton Event Generation on HPC Systems

This project explores the use of agentic systems to improve the efficiency of parton event generation workflows running on leadership-class high performance computing systems at the Argonne Leadership Computing Facility (ALCF). Parton event generators are a key component of high energy physics simulation pipelines, but they can require substantial data movement and bandwidth when executed at scale. The goal of this work is to investigate how agent-based orchestration and decision-making can be used to reduce bandwidth requirements and improve the efficiency of these workflows on large-scale computing systems. The project includes the design and evaluation of an agentic workflow capable of coordinating simulation components and managing data movement during execution. By incorporating adaptive control mechanisms, the system aims to optimize how tasks are scheduled and how intermediate data is communicated between stages of the workflow. This approach has the potential to reduce communication overhead while maintaining the performance and scientific fidelity required for large-scale physics simulations. Initial work has focused on surveying research literature on agentic systems and high performance computing workflows and developing the conceptual framework for implementing the approach on ALCF systems. The implementation and evaluation of the workflow will be carried out on Argonne supercomputing resources, with the goal of demonstrating how agent-based coordination can improve efficiency in HPC-based particle physics simulation pipelines.