UMD Graduate Student Daniel Nichols Awarded 2024 ACM-IEEE CS George Michael Memorial HPC Fellowship

University of Maryland Department of Computer Science graduate student Daniel Nichols has been named the recipient of the 2024 ACM-IEEE CS George Michael Memorial High-Performance Computing (HPC) Fellowship. This prestigious award acknowledges Nichols' research at the intersection of machine learning and high-performance computing, specifically in advancing machine-learning-based performance modeling and adapting large language models (LLMs) for HPC applications.

The George Michael Memorial HPC Fellowship, established in honor of George Michael—one of the founders of the SC Conference series—recognizes exceptional Ph.D. students globally whose research centers on high-performance computing applications, networking, storage, or large-scale data analytics. The fellowship includes a $5,000 honorarium and covers travel expenses for recipients to attend the annual SC conference, where the award is formally presented.

Nichols’ research addresses key challenges in performance modeling within the HPC domain, an area that is becoming increasingly crucial as computational demands continue to grow. His work focuses on developing machine-learning-based performance models that utilize all available performance data when predicting code runtime properties. Traditional performance models in HPC often rely on limited datasets, which can lead to inaccuracies and inefficiencies. By leveraging advancements in representation learning, Nichols aims to create models that offer more comprehensive and reliable predictions, ultimately enhancing the efficiency and applicability of performance models in HPC environments.

In addition to his work on performance modeling, Nichols has made significant contributions to the application of large language models in HPC. His research in this area seeks to adapt state-of-the-art LLM techniques to meet the specific needs of HPC applications, which often involve complex, scientific and parallel code. The challenges in this domain are substantial, as LLMs typically excel in general-purpose language tasks but require significant adaptation to handle the intricacies of HPC code.

Nichols’ approach involves creating LLMs that are specifically tailored to the unique demands of scientific and parallel computing. These specialized models aim to improve the performance of HPC development tasks, allowing researchers and scientists to focus more on their domain-specific research rather than the complexities of HPC development. 

About ACM

The Association for Computing Machinery (ACM) is the world’s largest educational and scientific computing society, bringing together computing educators, researchers and professionals to inspire dialogue, share resources and address challenges in the field. ACM amplifies the collective voice of the computing profession through leadership, promotion of high standards and recognition of technical excellence. The organization supports its members' professional growth by offering opportunities for lifelong learning, career development and networking.

—By Samuel Malede Zewdu, CS Communications

—Adapted from a press release by ACM