AI on HPC

The rapid advancement of Artificial Intelligence (AI) and Machine Learning (ML) has positioned High-Performance Computing (HPC) systems as indispensable platforms for developing, training, and executing these workloads. However, the architectural complexity and batch-oriented design of traditional HPC systems pose unique challenges distinct from those encountered in resource-elastic environments such as clouds. As AI models, such as large language, vision foundation, and multimodal architecture continue to grow in scale, their performance and scalability increasingly depend on HPC-grade architectures, advanced networking, and software engineering practices. The emergence of measurement-driven performance engineering and hybrid AI/HPC workflows is reshaping how large-scale computation is designed, benchmarked, and optimized.

The parallelization characteristics, input/output requirements, and dynamic workflows of AI workloads demand innovative techniques for efficient utilization of HPC resources. Moreover, the performance engineering of such workloads is crucial to achieve scalability, portability, and reproducibility across diverse system architectures.

We invite submissions presenting experimental results, architectural insights, performance studies, and best practices related to AI/ML workloads on HPC systems.

• → Full papers should be 4–10 pages in two-column (IEEE style).
• → Lightning talks: 1-page abstract (excluded from proceedings).
• → All submissions should include an artifact description appendix (AD).
• → Participating in the artifact evaluation process (AE) is encouraged.
• → Bibliography, AD and AE appendices do not count against the page limit.
• → The review process is double-anonymous (double-blind).
• → Workshop papers are planned to be published in the SC Workshops Proceedings volume.
• → "For an accepted paper to be included in the proceedings, the paper has to be presented at the conference in person." [SC26]