TITLE: From Algorithms to Communities in Physical Prediction, Inference, and Design
PRESENTER: Dr. Jed Brown, University of Colorado Boulder
DATE: Monday, January 30, 2023
TIME: 1:00 p.m. – 2:00 p.m.
LOCATION: Comal 212
ABSTRACT:
Modeling and simulation are essential tools for modern engineering and natural sciences, leveraging domain knowledge and indirect data to create reliable causal models that can infer latent
variables and optimize new products and experiments. Such models are often expressed in terms of partial differential equations and practitioners have an insatiable thirst for large-scale computation to
provide sufficient accuracy, even for deterministic predictive simulation. I'll discuss ways in which modern architectures, including GPUs, have required transformational changes to data structures and
abstractions, with some examples of our work in this area and deployment through community software libraries. I'll offer a view into experiences building sustainable multi-stakeholder communities
around these libraries, which are foundational to simulation efforts in industry, national labs, and academia. Finally, I'll discuss academic incentives, the Journal of Open-Source Software, and a vision for expanding community software development principles and methodology to lower barriers, increase research quality, and encourage productive collaboration.
BIO:
Jed was raised by wolves in a remote part of interior Alaska. Not surprisingly, his high-school mascot was a Wolfpack, but he defected in college and earned All American honors in cross-country skiing while racing for the Nanooks (Inuit for polar bear) at the University of Alaska Fairbanks. Always fond of cold places, he wrote the Parallel Ice Sheet Model (PISM) while at UAF, and later earned a DrSc at ETH Zürich investigating computational methods for ice sheet and glacier dynamics. Gradually warming up to temperate climates, he won the 2009 Piolet d'Or climbing award for the first ascent of Xuelian West in China's Tien Shan and has subsequently solved high-temperature problems while at Argonne National Lab. He has been a developer of the Portable Extensible Toolkit for Scientific computing
(PETSc) since 2008 and is the creator of libCEED, a performance-portable library for element-based algebra. He was Asst. Computational Mathematician at Argonne prior to joining the faculty at the University of Colorado Boulder in 2015. His work has been recognized by the 2014 SIAG/SC Junior Scientist Prize and a 2014 IEEE TCSC Young Achiever Award, and as co-recipient of the 2015 SIAM/ACM Prize in Computational Science and Engineering.
