We develop scalable algorithms for quantum and classical molecular simulations, achieving world-record performance on the largest supercomputers in the world, such as the first exascale machine Frontier, leveraging the computational power of over 37,000 GPUs. Our EXESS software achieves exaflop-scale quantum simulations across thousands of GPUs, enabling quantum chemical accuracy for systems previously out of reach, such as protein–ligand complexes and covalent inhibitors.
Key methods include:
Impact: Enables virtual testing of large therapeutic candidates with quantum-level precision.
Our AI research develops physics-informed machine learning to predict molecular energetics and properties with minimal data. Using equivariant neural networks, Bayesian models, and generative AI, we:
Impact: AI tools trained on our quantum simulation data reduce the need for expensive physical screens, and enable novel, previously inaccessible therapeutic designs.
We engineer modular and automated workflows that mimic the full drug discovery pipeline—from hit identification to quantum-level optimization and final lead selection. These pipelines integrate:
Impact: A fully autonomous drug discovery lab for therapeutic innovation would accelerate by order of magnitude the genereationa and testing of novel therapeutics.