معرفی
Joseph Devietti is an Associate Professor in the Department of Computer & Information Science at the University of Pennsylvania. His research focuses on improving programmability and performance of multiprocessor systems through architectural and programming model innovations. He actively advises PhD students and has supervised numerous graduates now employed at leading tech companies and academic institutions.
- Education: PhD (2012), MS (2009) in Computer Science and Engineering from University of Washington; BSE (2006) in Computer Science and BA (2006) in English from University of Pennsylvania.
- Employment: Associate Professor (2020–present), Assistant Professor (2013–2020) at University of Pennsylvania; Principal Scientist & Co-founder at Cloudseal, Inc. (2018–2020).
Devietti’s research spans computer architecture, parallel programming, and deterministic execution. Key areas include cache/memory optimization (prefetching, false sharing repair), GPU programming models (race detection, block-size independence), and hardware-software co-design for concurrency safety. His recent work addresses dynamic runtime prefetch tuning (RPG2), online code layout optimization (OCOLOS), and intelligent BTB prefetching (Twig) for data center applications.
His publications from 2024–2017 reveal trends in instruction/cache optimization (2024–2020), GPU determinism (2018–2017), and race detection (2018–2016). Awards include the 2024 Penn Engineering Ford Motor Company Award, Radhia Cousot Best Paper (2018), and IEEE Micro Top Picks recognition (2023, 2009, 2008).
- Scientific Awards:
- 2024 Penn Engineering Ford Motor Company Award
- Radhia Cousot Young Researcher Best Paper Award (SAS 2018)
- IEEE Micro Top Picks (2023, 2009, 2008)
- Intel Early Career Faculty Honor Program (2013)
- Intel Ph.D. Fellowship (2011)
- Advising: Supervised 15+ PhD/Master’s students with placements at Google, Microsoft, Amazon, NYU, and the United States Naval Academy.
- Collaborations: Works with industry leaders (NVIDIA, Facebook) and academic institutions (University of Washington, Penn).
