We develop domain-informed AI methods for chemical engineering, molecular, process, and manufacturing systems.
Industrial data are rarely clean or complete. We develop practical AI pipelines with knowledge for robust prediction, missing-value imputation, explainable AI, and inverse design.
We build fast and reliable models for monitoring, control, and optimization. Current work includes reduced-order modeling and digital-twin pipelines for battery-electrode manufacturing, alongside broader applications.
We study representations and learning architectures for catalysts, solvents, absorbents, lubricants, and composites. Our focus is on models that can explore chemical space while remaining interpretable and physically meaningful.
We integrate single & multiobjective optimization for economic, environmental, and operational advantages.
Embed domain knowledge and physics so that models can be reliable beyond data.
Replace costly experiments and simulations with accurate surrogates that speed up screening and design.
Connect models with optimization to move from prediction toward autonomous engineering workflows.
