Ashna Nawar Ahmed is a Doctoral Research Assistant at Texas State University, specializing in surrogate models for high-performance computing (HPC) optimization, who recently had a paper accepted at the Neural Information Processing Systems (NeurIPS) 2025 conference and a poster accepted for presentation at the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC25). These successes were in large part due to her internship last summer at Oak Ridge National Laboratory (ORNL), which came about partly because of one seminar.
Ashna’s supervisor, Dr. Tanzima Islam, Associate Professor in the Department of Computer Science at Texas State University, encouraged her research group to attend “Multi-Output Surrogate Construction for Fusion Simulations” Dr. Kathryn Maupin, Senior Member of Technical Staff, Sandia National Laboratory. The seminar was part of the Computational Research Leadership Council (CRLC) seminar series and aligned closely with Ashna’s research.
“My work focuses on surrogate models for high-performance computing (HPC) optimization, so I was very interested in learning how similar techniques are applied in scientific simulation contexts,” says Ashna. “It offered useful perspectives on multi-output modeling and practical considerations for large-scale simulations. My participation in the seminar certainly also helped strengthen my understanding of national-lab research environments and topics that align with their work.”
In the summer of 2025, Ashna completed an internship with Terry Jones, Senior Research Staff (Computer Scientist) in the Computer Science and Mathematics Division at ORNL. This internship deepened her own research and ultimately led to the poster “Intelligent Surrogates Pay Attention to Data, Improving Multi-Objective HPC Optimization,” authored by Ashna, Dr. Islam, Mr. Jones, and Banooqa Banday, another doctoral student at Texas State University, which was presented at SC25. This poster presents their work on applying embedding-informed surrogate models and multi-objective Bayesian optimization to improve HPC scheduling.
Ashna’s internship also led to the paper “Attention-Informed Surrogates for Navigating Power–Performance Trade-offs in HPC,” authored by the same team, that was accepted to the Machine Learning for Systems (MLForSys) Workshop at NeurIPS 2025.
“I really appreciate the CRLC Seminar series. It has been an excellent way to learn about emerging topics and connect classroom or lab research to real-world scientific applications,” says Ashna. “The series also helped me think more deeply about potential extensions of my work.” And those extensions are now finding real-world success, helping open even more opportunities, including potentially a Sustainable Research Pathways for National Artificial Intelligence Research Resource (NAIRR) internship program, which she has applied for next year.