Sunita Chandrasekaran to Lead Team Preparing America’s Next-Gen 2029 Supercomputer 'Discovery'

Chandrasekaran, an associate professor of computer and information sciences at the University of Delaware, will coordinate a multi-institutional effort to evaluate the system's performance and readiness ahead of its deployment. The team will run a suite of benchmark tests and real-world scientific workloads to identify bottlenecks and optimize the machine's software stack.

Discovery, built by Hewlett Packard Enterprise, is expected to deliver more than five times the performance of Frontier, the current fastest supercomputer in the world, which also resides at Oak Ridge National Laboratory in Tennessee. The new machine will be capable of over 10 exaflops, or 10 quintillion calculations per second, and will be used for research in areas such as climate modeling, drug discovery, and materials science.

The University of Delaware team is one of several selected by the Department of Energy's Exascale Computing Project to participate in the early testing phase. Chandrasekaran's group will focus on evaluating the system's ability to handle AI-driven workloads, including large language models and scientific machine learning tasks, which are expected to be a major use case for Discovery.

Chandrasekaran, who holds a Ph.D. in computer science from the University of Houston, has previously worked on performance optimization for high-performance computing systems. Her research interests include parallel computing, programming models, and the intersection of AI and scientific computing.

The testing phase is scheduled to begin in 2027, with the system expected to be fully operational by early 2029. The findings from the stress tests will inform final adjustments to the hardware and software before Discovery opens to researchers nationwide.