Underwood honored for outstanding performance in basic research

Robert Underwood is a recipient of a 2024 Outstanding Postdoctoral Performance Award in Basic Research. The award was presented at the Postdoctoral Research and Career Symposium at Argonne on November 21, 2024.

After receiving his Ph.D. in computer science from Clemson University in 2021, Underwood joined Argonne as a postdoctoral appointee in the Mathematics and Computer Science Division; he was recently converted to assistant computer scientist in the division. 

“I am honored to receive this postdoctoral performance award from Argonne,” Underwood said. ​“I look forward to continuing my research looking at challenging problems in data and I/O with the fantastic community of scientists we have at Argonne.”

Underwood has made significant contributions to the development of data compression methods for large-scale applications such as light sources and climate. He is the developer of LibPressio, a library that abstracts between different lossless and lossy compressors, thus minimizing the code changes scientists need in order to take advantage of new and improved compression techniques while allowing compression developers to innovate more quickly. The library is used in over 17 institutions worldwide.

Another research area that Underwood is exploring at Argonne is development of large language models (LLMs) specialized for scientific use cases. He is the co-lead for the Data team of AuroraGPT, which aims to leverage U.S. Department of Energy supercomputing resources to develop and enhance understanding of LLMs for science. He gave a workshop presentation at the SC24 conference discussing the team’s progress in preparing terabytes of scientific data and text for training LLMs. 

Underwood also has been involved in mentoring activities. He has been the Early Career Ambassador for Argonne, helping connect new students and postdocs at Argonne to the Joint Laboratory for Extreme-Scale Computing and representing their needs to the joint lab.