307 Rice Hall
85 Engineer's Way
Charlottesville, VA 22904
I am an associate professor of Computer Science at the University of Virginia. I lead the Responsible AI for Science and Engineering (RAISE) group, where we develop foundational machine learning and generative AI methods that make models obey control objectives, hard constraints, safety requirements, and physical laws. Our work studies how to build controllable foundation models and constraint-aware generative models by connecting learning with optimization, control, and neuro-symbolic reasoning. My research has received several honors, including best paper awards at leading journals and AI/ML conferences, industry research awards, fellowships, and young investigator awards. See the awards page for more details.
Research Focus
- Foundational ML and Generative AI under Constraints Differentiable optimization and the integration of optimization with machine learning and generative AI.
- AI for Science and EngineeringGenerative methods for protein and molecular design, materials science, robotics, energy systems, chip design, and policy optimization.
- Discrete Diffusion Language ModelsFoundations for constrained discrete generation and its use in speculative technologies for speeding up language generation.
- Decision Focused LearningTraining predictive and generative models to optimize downstream decision quality, rather than prediction accuracy alone.
- LLM Multiagent SystemsFoundations for distributed coordination and problem solving and Safety
- Responsible AIAssurance for learning and decision systems, with an emphasis on privacy, safety, fairness, and robustness.
Research Sponsors
Our group is grateful for the generous support of our sponsors.
Highlighted Work
Constraint-Aware Generative AI for Science and Engineering
How RAISE Lab makes generative models obey physics, geometry, logic, and safety requirements.
Learn more
Constrained Discrete Diffusion for Language, Chemistry, and Code
Enforcing hard constraints on tokens, molecules, and programs — training-free, inside the denoising loop.
Learn more
Massively Speeding up LLM inference with Speculative Diffusion Decoding
Faster LLM inference with discrete diffusion drafting and alignment.
Learn moreRecent News
40 total-
May 2026
Publication- Our paper Stability-Constrained AC Optimal Power Flow–A Gaussian Process-Based Approach has been accepted to Sustainable Energy, Grids and Networks. Congrats to Vincenzo!
- New preprints:- Simulation-Informed Diffusion for Decentralized Multi-robot Motion Planning
- Constraint-Aware Flow Matching: Decision Aligned End-to-End Training for Constrained Sampling
- Constrained Code Generation with Discrete Diffusion
- Simple Self-Conditioning Adaptation for Masked Diffusion Models
- I’ll be teaching a course on constrained-aware generative AI at the AI-SCORE summer school in Chicago later this month. -
Mar 2026
Publication- Our paper Learning to Optimize meets Neural-ODE: Real-Time, Stability-Constrained AC OPF has been accepted in the Electric Power Systems Research journal. Congratulations to Vincenzo and Mostafa!
- Together with Kyusang Lee and Hong Joo Moon, we received a 3Cavaliers Funding Award on AI-enabled implantable sensors for personalized monitoring after spinal surgery.
-
Feb 2026
Publication- New preprint on multi-agent LLM systems and the emergence of collusive behaviors: Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems.
- New preprint on Masked Diffusion Models + Constraints! The idea uses search augmentation during the reverse denoising step of diffusion to optimize for constraint violations: Search-Augmented Masked Diffusion Models for Constrained Generation.
-
Jan 2026
Publication- Two ICLR 2026 papers accepted, on Constrained Diffusion for Protein Design and on Decision-Focused Generative Learning. Congrats to Jacob and Jinhao and co-authors!
- Our SpecDiff-2 work was accepted at MLSys 2026! Congrats to Jameson, Jacob, and co-authors!
- Paper on Learning to optimize in Bilevel Optimization accepted to L4DC 2026. Congrats to James and co-authors!
- New preprint on privacy guardrails: NeuroFilter: Privacy Guardrails for Conversational LLM Agents.
-
Dec 2025
Publication- Paper accepted at IEEE Transactions on Smart Grid on decision-focused learning with neural ODEs for proactive grid resilience management.
- Paper accepted at IEEE SaTML 2026 on data minimization.
- Our work Constrained Molecular Generation with Discrete Diffusion received the Best Student Paper Award at the NeurIPS AI4D3 Workshop! Congrats to Jacob and Michael!
- I gave a talk at the NeurIPS COML Workshop on Generative AI for Scientific and Engineering Research.
-
Nov 2025
Publication- Our work Discrete-Guided Diffusion for Scalable and Safe Multi-Robot Motion Planning was accepted at AAAI 2026! Congrats to Jinhao!
- Check out our new work on accelerating LLM inference with discrete diffusion models! These results are amazing!
- I gave a talk at the ORNL AI Core workshop on Generative AI for Science.
-
Oct 2025
Publication- Check out our new work on manifold constrained protein design with diffusion models!.
- Several new preprints on topics spanning from physics-constrained generative flow matching and diffusion models to safety in agentic AI systems and learning to solve PDE-constrained problems.
Check out our publications page for details.
-
Sep 2025
Publication- Two papers on physics-constrained generative diffusion models (as Spotlight) and logic-constrained language diffusion models accepted at NeurIPS 2025
Congratulations to Jacob, Michael, and Stefano!
- Our book Differential Privacy in Artificial Intelligence: From Theory to Practice is now online, available as Open Access!
This book covers the (i) theoretical underpinnings of Differential Privacy, (ii) recent algorithmic advances for machine learning,, (iii) practical applications across key engineering and scientific domains, (iv) methodologies for implementation and empirical evaluation, and (v) the surrounding legal and ethical frameworks.
The book is also available on Amazon and Barnes & Noble. -
Aug 2025
Publication- New preprint: Discrete-Guided Diffusion for Scalable and Safe Multi-Robot Motion Planning. We provide a new integration of discrete optimization and diffusion models for multi-robot motion planning, scaling to 100 agents in cluttered environments.
- New preprint: Stability-Constrained AC Optimal Power Flow–A Gaussian Process-Based Approach. We propose a new, scalable approach to incorporate generator dynamics into the ACOPF using Gaussian Process (GP) models. Crucially, this enables probabilistic stability assessment to be integrated directly into the optimization process!
- NSF-RI proposal on Generative Models for Scientific Exploration funded! We’ll integrate optimization and physical principles directly in generative models for a variety of scientific applications! Thank you, NSF!
- Paper Privacy-Preserving Convex Optimization: When Differential Privacy Meets Stochastic Programming accepted at CDC-2025!
- New preprint: SoK: Data Minimization in Machine Learning. We provide introduces a comprehensive framework for data minimization in machine learning and introduce a systematical review the literature on data minimization and adjacent methodologies.
-
Jul 2025
Award- Congratulations to Michael Cardei for winning an NSF GRFP Scholarship Award!
- I gave a talk at USNCCM18 about our generative AI for material discovery.