Kevin Pedro
Fermi National Accelerator Laboratory (FNAL)
Biography
Kevin Pedro received his bachelor’s degree in physics from Rensselaer Polytechnic Institute and his PhD from the University of Maryland. He is a scientist at Fermi National Accelerator Laboratory and a member of the CMS collaboration, which he first joined in 2009. He has also worked on Daya Bay, ATLAS, CLIC, DUNE, astrophysics, and phenomenology. He is currently the CMS detector simulation convener and has previously held numerous other leadership roles in CMS software management and the broader community, including the HEP Software Foundation. He leads the search program for signatures of composite dark matter from dark QCD models, with phenomenological signatures including semivisible jets, emerging jets, and soft unclustered energy patterns. His technical research focuses on the intersection of machine learning and scientific computing, including inference as a service, generative ML for simulation, and unsupervised ML for model-independent searches for physics beyond the standard model.
Project interests
- Machine learning for detector simulation, pileup mixing, and other high-multiplicity background overlay problems. Kevin is the principal investigator for the CaloDiffusion, a leading generative ML approach.
- Improving classical fast simulation techniques using parameterized approaches and/or GPU simulation engines.
- Expanding inference as a service for flexible heterogeneous computing in experiment software frameworks, such as the services for optimized network inference on coprocessors (SONIC), for which he is the lead developer in CMS.
- Software design and simulations for future collider experiments, primarily the proposed 10 TeV Muon Collider.