معرفی
Edgar Dobriban is an Associate Professor of Statistics and Data Science at the University of Pennsylvania's Wharton School, with a secondary appointment in Computer and Information Science. He leads a research group focused on problems at the interface of statistics, machine learning, and AI.
Education
- PhD in Statistics, Stanford University (2017)
- BA in Mathematics, Princeton University (2012, Summa cum Laude/with Highest Honors)
Research Focus
His work spans uncertainty quantification, AI safety, robustness, high-dimensional asymptotic statistics, distributed learning, fairness, and COVID-19 testing methodologies. Current projects include developing conformal prediction methods, jailbreaking robustness benchmarks (JailbreakBench), and safety alignment techniques for large language models.
Publication Trends
Recent papers predominantly address AI safety and reliability, featuring novel methods for uncertainty quantification in language models (calibration, conformal prediction), adversarial robustness (jailbreaking defenses), and distribution shift adaptation. Theoretical foundations blend with practical applications in high-dimensional statistics.
Awards and Honors
- Peter Gavin Hall IMS Early Career Prize (2024)
- Sloan Research Fellowship (2023)
- ICSA Outstanding Young Researcher Award (2023)
- NSF CAREER Award (2021)
- AFOSR/Army Research Office YIP Awards (2024, 2023)
- COPSS Emerging Leader Award (2023)
Research Leadership
He leads the Wharton Statistics and Data Science research group, recruiting PhD students through Statistics & Data Science, CIS, and AMCS programs. Current projects involve collaborations with Penn Medicine and the NSF-Simons Mathematical and Scientific Foundations of Deep Learning initiative. He co-founded the ASA StatsUpAI Special Interest Group and co-organized the Shenzhen Conference on Random Matrix Theory (2023).
