The overarching goal of our research is to develop theory, algorithms, models, and software for large-scale optimization and machine learning with applications in energy, chemical, and biological systems.

• Novel Deep Learning Models Informed by Physics and Domain Knowledge
  Deep neural networks have transformed engineering but often act as “black boxes,” producing results that may violate physical laws in safety-critical applications. We develop optimization-inspired neural networks (OINNs) that rigorously enforce physical and logical constraints, creating reliable and interpretable models for process design and control.

• Explain Optimization and Machine Learning Models Using Generative AI
  Optimization and machine learning models are powerful but inaccessible to many non-technical stakeholders, creating a “language gap” that hinders collaboration and trust. We build large language model (LLM)-powered interfaces that allow users to ask questions, diagnose issues, and interpret model results in plain language.

• Machine Learning for Discrete and Global Optimization
  Many industrial decision problems involve complex mixed-integer nonlinear programming (MINLP), where traditional solvers are too slow and machine learning lacks optimality guarantees. We combine deep learning and reinforcement learning with optimization techniques to accelerate solution processes while maintaining reliability.

• Data Sharing for Decarbonization
  Industrial ecosystems miss opportunities for system-wide CO reduction because stakeholders optimize locally while keeping data confidential. We create secure, federated data-sharing frameworks that protect sensitive information while enabling joint optimization of energy use and emissions.