AI for the Physical World

Foundation models have transformed how machines reason about language, but the next frontier is enabling AI to reason about the physical world. Our research focuses on developing the data, workflows, and computational infrastructure that allow AI systems to understand molecules, materials, biological systems, and medicine. By grounding AI in scientific principles and experimental evidence, we aim to accelerate discovery across the natural sciences.


Agentic Scientific Intelligence

Scientific discovery rarely depends on a single model. It requires planning, simulation, experimentation, validation, and iterative refinement. We develop multi-agent systems that orchestrate specialized models, scientific software, laboratory automation, and human expertise into cohesive workflows capable of solving complex scientific problems. Our goal is to transform frontier models into reliable scientific collaborators rather than standalone reasoning engines.


Scientific Infrastructure

Progress in scientific AI depends on more than increasingly capable foundation models. It requires high-quality datasets, rigorous evaluation, specialized tools, and reproducible workflows. We build reusable infrastructure—including benchmark datasets, evaluation harnesses, simulation interfaces, code-execution environments, and domain-specific toolboxes—that enables frontier AI systems to perform trustworthy scientific reasoning across chemistry, biology, materials science, and medicine.


Physics- and Biology-Grounded AI

Scientific AI must respect the constraints of the real world. Our research integrates physical laws, chemical principles, biological mechanisms, and experimental observations directly into AI workflows. We combine computational models with simulations, laboratory instrumentation, multimodal scientific data, and mechanistic validation to produce systems that are interpretable, reliable, and experimentally meaningful.


Collaborative Research

The most impactful scientific advances emerge through collaboration. We partner with frontier AI organizations, research institutions, and industry to tackle challenging problems in molecular discovery, materials science, precision medicine, and Physical AI. By combining complementary expertise with reusable scientific infrastructure, we seek to accelerate discovery while advancing the broader scientific AI ecosystem.