Scholar
Gagandeep Singh
Google Scholar ID: m4b2ruEAAAAJ
Assistant Professor, Department of Computer Science, UIUC
Citations & Impact
Citations
3,537
H-index
24
i10-index
38
Publications
20
Co-authors
76
list available
Publications
Resume (English only)
Academic Achievements
- Co-recipient of the ACM SIGPLAN Doctoral Dissertation Award, given annually to the best dissertations in the area of Programming Languages. Published multiple academic papers, including a monograph on applying the classical framework of Abstract Interpretation for AI safety in Foundations and Trends in Programming Languages. Recent research achievements include the first compression-aware certified training framework, a novel theoretical study of CoT reasoning, a new efficient RLHF algorithm that does not require KL regularization, efficient adversarial attacks on agentic systems, the first algorithm for constrained generation with diffusion LLMs, and new methods that reveal misalignment between CoT reasoning and the final answer.
Research Experience
- During his PhD, he designed scalable and precise automated reasoning methods and tools for programs and deep neural networks. Currently an Assistant Professor at UIUC, leading the FOCAL Lab.
Education
- PhD in Computer Science from ETH Zurich in 2020, supervised by Prof. Markus Püschel and Prof. Martin Vechev. Master's in Computer Science from ETH Zurich in 2014, receiving the ETH Master Medal. Bachelor's in Computer Science and Engineering from IIT Patna in 2012, receiving the President of India Gold Medal.
Background
- Currently an Assistant Professor in the Department of Computer Science at the University of Illinois Urbana-Champaign (UIUC), leading the FOrmally Certified Automation and Learning (FOCAL) Lab. Research interests include combining ideas from Formal Logic, Machine Learning, and Systems research to develop systematic and theoretically principled approaches for intelligent compute systems with formal guarantees about their behavior and safety. Current research topics include: Formal Methods for LLMs, Robust and Efficient Reinforcement Learning, Learning from Preference Data, LLMs for Mathematical Reasoning, Theoretical Understanding of LLMs, Securing Agentic AI, Neural Networks for Computer Systems with Formal Guarantees, Provably Robust Machine Learning for Wireless, Explainable AI, Neural Network Verification, Training with Logical Constraints, NeuroSymbolic Program Analysis, Numerical Abstract Interpretation.
Miscellany
- Interested in recruiting PhD candidates, particularly those interested in the intersection of AI, Social Sciences (e.g., Economics), and Policy Making.
