Research Overview
Our research at SECQUOIA (Systems Engineering, Computation, and Quantum Optimization for AI) spans multiple areas at the intersection of optimization, artificial intelligence, and emerging computation paradigms. We develop theory, algorithms, and software for large-scale optimization, with applications in process systems engineering, data-driven decision-making, and quantum-enhanced computation. Our work is problem-driven and interdisciplinary, with active collaborations in chemical engineering, quantum information, and computer science.
Generalized Disjunctive Programming and Mixed-Integer Nonlinear Optimization
We develop exact and approximation algorithms for solving generalized disjunctive programming (GDP) models arising in process network synthesis, scheduling, and design under uncertainty. Current work explores tight reformulations, surrogate-based relaxations, and scalable solution strategies for nonlinear and dynamic systems with discrete decisions.
Quantum Optimization and Quantum-Inspired Algorithms
We study how quantum computing can accelerate optimization by designing and benchmarking hybrid quantum-classical algorithms. This includes variational quantum algorithms for discrete-continuous problems, quantum-enhanced surrogate modeling, and tensor network simulations to evaluate algorithmic performance under realistic noise models.
Federated Learning for Process and Biomedical Applications
Our group explores federated learning in settings where data privacy, heterogeneity, and limited communication are key constraints. Applications include predictive modeling in pharmaceutical manufacturing and multimodal learning from DNA and MRI data for disease classification. We also investigate secure federated learning using fully homomorphic encryption.
AI-Augmented Process Engineering
We integrate machine learning into process simulation and control workflows by leveraging domain knowledge and physical models. Projects include developing digital twins, interpretable surrogate models, and adaptive model-predictive control for chemical and energy systems.
Open-Source Software for Optimization and AI
We contribute to the development of open-source tools to advance reproducibility and usability of optimization and learning methods. Ongoing efforts include software for mixed-integer optimization, simulation-based optimization, federated learning, and quantum algorithm simulation.