Research

We develop theory, convexification and hull reformulations, decomposition algorithms, and derivative-free configuration search methods for Generalized Disjunctive Programming (GDP) and Mixed-Integer Nonlinear Programming (MINLP). Representative directions include regularized outer approximation, logic-based discrete steepest descent, quadratic disjunctive programming, and hybrid branch-and-bound methods.

Our optimization frameworks address core synthesis, design, and operational challenges in process systems engineering—spanning catalytic distillation and process intensification, reverse electrodialysis and water-energy networks, refinery scheduling, bio-based chemical manufacturing, and resilient supply chains.

We investigate how gate-model quantum processors, quantum annealers, coherent Ising machines, and GPU-accelerated tensor networks tackle structured discrete optimization and Hamiltonian simulation problems. Work spans hybrid quantum-classical decomposition, QUBO/Ising encodings, circuit transpilation, and utility-scale hardware evaluation.

We design machine learning and AI methodologies that respect physical constraints, distributed data ownership, and scientific verification—covering privacy-preserving federated and quantum-federated learning across chemical engineering and healthcare, surrogate/Bayesian circuit optimization, and multi-agent AI research workflows.

Reproducible open-source software is a first-class scholarly output of our group. We build and maintain algebraic modeling extensions, MINLP/GDP decomposition solvers, problem reformulation compilers, and community benchmark suites across the Python (Pyomo) and Julia (JuMP / JuliaQUBO) ecosystems.