Workshop: AgenticAI4HPC’26: The First International Workshop on Agentic AI for HPC
Authors: Subrahmanya Vinayak Joshi, Aalap Tripathy, Scott Hinchley, Deepak Nanjundaiah, Sergey Serebryakov, Shreni Agrawal, Gayathri Saranathan, Tarun Kumar, Don Bahls, Charlie Vollmer, Robert Wisniewski, Jeff Hanson, Tejas Hegde, and Prince Jain (HPE)
Abstract: High-performance computing (HPC) systems are large, complex environments comprising thousands of compute nodes, high-speed networks, storage systems, and dedicated management infrastructure. Despite the availability of advanced monitoring and management tools, users and administrators often face challenges due to the volume of operational data, the complexity of system interactions, and the expertise required to effectively interpret and act on available information. Recent advances in agentic AI (artificial intelligence) offer new opportunities for more autonomous, context-aware, and collaborative system management. In this work, we present an HPC Systems Management Copilot that integrates a collection of intelligent agents capable of interacting with job schedulers, monitoring frameworks, system databases, and other HPC management services. The copilot provides real-time assistance to users and administrators through a unified natural-language interface. We also present an experimental evaluation assessing the capabilities, effectiveness, and safety of the proposed system across representative HPC management scenarios.
Back to AgenticAI4HPC’26: The First International Workshop on Agentic AI for HPC Archive Listing
Back to Full Workshop Archive Listing