My group develops algorithms and AI systems that combine mathematical structure, scalable learning, and domain knowledge for high-impact scientific and industrial problems.
Scalable optimization methods for machine learning, including stochastic, distributed, adaptive, and second-order algorithms for modern AI systems.
Learning systems for scientific discovery, with emphasis on structured prediction, biology, causal modeling, and models that respect scientific constraints.
Domain-specific LLMs, VLMs, and reasoning systems for exploration, production, subsurface workflows, uncertainty, and decision support in energy.
Reasoning, prompting, multi-agent systems, evaluation, and domain adaptation for language models that need to be reliable in specialized workflows.
Across these directions, the recurring theme is rigorous learning under constraints: algorithms that scale, models that reason, and systems that can be deployed in domains where correctness, uncertainty, and scientific structure matter.
Browse publications View flagship projects