Workshop: High Performance Python for Science at Scale
Authors: Pete Mendygral (HPE), Rafael Ferreira da Silva (Oak Ridge National Laboratory)
Abstract: The DOE Genesis Mission aims to couple AI, exascale computing, experimental facilities, and autonomous laboratories into an integrated discovery platform, and much of what holds that platform together is written in Python. Large-scale simulation campaigns, AI training and inference pipelines, workflow orchestration, provenance capture, and multi-agent systems all converge on a shared Python layer that lets scientists compose resources across institutional and facility boundaries. Drawing on experience designing agentic systems for scientific resource discovery and workflow composition, performance studies of workflow execution on leadership-class machines, and a decade of community building around scientific workflows, this talk examines where Python succeeds as the orchestration fabric for AI-coupled science and where it strains, from startup and coordination overheads at scale to environment portability across facilities, concurrency, and the cost of provenance and governance in autonomous loops.
Back to High Performance Python for Science at Scale Archive Listing
Back to Full Workshop Archive Listing