Scholar
Jae-Won Chung
Google Scholar ID: xkSoEDYAAAAJ
University of Michigan
Citations & Impact
Citations
409
H-index
5
i10-index
4
Publications
9
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- Paper: 'The ML.ENERGY Benchmark: Toward Automated Inference Energy Measurement and Optimization', NeurIPS D&B spotlight, 2025, Spotlight acceptance rate = 2.81%.
- Paper: 'Perseus: Reducing Energy Bloat in Large Model Training', SOSP, 2024, Acceptance rate = 17.34%.
- Paper: 'Toward Cross-Layer Energy Optimizations in AI Systems', DOE ASCR Energy-Efficient Computing for Science Workshop, 2024.
- Preprint: 'Andes: Defining and Enhancing Quality-of-Experience in LLM-Based Text Streaming Services', 2024.
- Paper: 'Zeus: Understanding and Optimizing GPU Energy Consumption of DNN Training', USENIX NSDI, 2023, Acceptance rate = 18.38%.
Research Experience
- Graduate Student Research Assistant, SymbioticLab, UMich, Sep 2021 - Expected May 2027, Advisor: Prof. Mosharaf Chowdhury, Building software systems for machine learning that treat power and energy as first-class systems resources.
- Research Scientist Intern, AI and Systems Co-Design Team, Meta, May 2025 - Aug 2025, Supporting MoE training on MTIA platforms.
- Research Intern, Software Platform Lab, SNU, Mar 2020 - Jun 2021, Advisor: Prof. Byung-Gon Chun, Developed Crane, a GPU cluster manager for elastic AutoML jobs.
- Research Intern, Virtual Machine and Optimization Lab, SNU, Dec 2019 - Jun 2020, Advisor: Prof. Soo-Mook Moon, Created ShadowTutor, a server-client collaborative DNN inference system.
- Research Intern, Computer Vision Lab, SNU, Jun 2019 - Dec 2019, Advisor: Prof. Kyoung Mu Lee, Worked on finding better meta-initialization points for Model-Agnostic Meta-Learning (MAML) using LSTM-based neural memory modules.
Education
- Ph.D. Candidate, Computer Science and Engineering, University of Michigan, advised by Professor Mosharaf Chowdhury, expected to graduate in May 2027.
Background
- Fifth year PhD candidate in CSE at the University of Michigan. Research interests include building efficient software systems for deep learning, with a focus on optimizing time and energy management. Views power and energy as fundamental systems resources worth carefully optimizing and allocating, both in hardware and from software.
Miscellany
- Leads the ML.ENERGY initiative as part of his research and open-source efforts.
