AMD unveils the open-source HPC Assistant, an AI tool to streamline supercomputing workflows, debuting on LUMI and Frontier.
AMD has introduced the AMD HPC Assistant, an open-source, AI-powered tool designed to enhance productivity for researchers using high-performance computing (HPC) systems. The assistant debuted on two of the world’s most advanced supercomputers: LUMI in Finland and Frontier in the United States. By leveraging agentic AI, the tool aims to simplify complex workflows, reduce learning curves, and optimize scientific computing processes.
The AMD HPC Assistant integrates natural language processing with HPC-specific knowledge, enabling researchers to interact with their computing environment more intuitively. Instead of navigating through documentation, support portals, or command-line interfaces, users can describe their objectives in plain language. The assistant then translates these inputs into actionable commands, such as generating job scripts, monitoring workloads, or diagnosing performance bottlenecks.
For example, researchers can ask where to store a 100 GB dataset, and the assistant will provide guidance based on the system’s storage policies, quotas, and purge rules. It also supports file synchronization, job submission, and GPU performance profiling, offering actionable insights into compute utilization, cache efficiency, and energy consumption.
The first implementation of the AMD HPC Assistant was developed in collaboration with CSC, the Finnish IT Center for Science, which operates LUMI, one of Europe’s most powerful supercomputers. The assistant is integrated with LUMI’s knowledge base, providing site-specific guidance tailored to its unique infrastructure and policies.
Building on this success, AMD is working with Oak Ridge National Laboratory to deploy the assistant on Frontier, the world’s first exascale supercomputer.
Source link







