Scholar
Tian Li
Google Scholar ID: 8JWoJrAAAAAJ
University of Chicago
Citations & Impact
Citations
21,331
H-index
20
i10-index
22
Publications
20
Co-authors
30
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Multiple ICML 2025 acceptances: BiClip optimizer, Tilted Sharpness-Aware Minimization, Exponential Tilting Generalization
- Work on many-objective optimization accepted to ICLR 2025 main conference; workshop papers on optimization under heavy-tailed noise and private RAG
- Publication Co-Chair for MLSys 2025
- Selected for AAAI New Faculty Highlights 2025 and gave a talk at AAAI
- Invited to serve on an NSF panel and as jury member for Germany's Composite Learning Challenge
- Received cloud credits from the NAIRR Pilot program in 2024
- Published 'Private Zeroth-Order Optimization with Public Data' and 'Efficient Adaptive Federated Optimization' at NeurIPS 2025
- Published 'Efficient Distributed Optimization under Heavy-Tailed Noise' and 'Tilted Sharpness-Aware Minimization' at ICML 2025
Background
- Assistant Professor at the Department of Computer Science and the Data Science Institute, University of Chicago
- Member of the UChicago Committee on Computational and Applied Mathematics
- Research focuses on large-scale machine learning and optimization
- Studies tradeoffs between model utility/convergence and system efficiency (memory, compute, communication costs)
- Improves these tradeoffs by designing cheaper optimizers, exploring new distributed training algorithms, and better utilizing data from diverse distributions
- Also investigates critical aspects beyond accuracy and efficiency, such as privacy and robustness in ML training and deployment
- Proposes practical definitions of privacy and robustness, studies interconnections among privacy, robustness, generalization, memorization, and reasoning, and designs provable scalable algorithms
