My Vision
My research in process systems engineering develops computational methods in machine learning and optimization to tackle environmental, societal, and sustainability challenges in engineering and science. Furthermore, I operate at the intersection of chemical engineering, mathematics/statistics, and computer science to drive an open-source software-accelerated research paradigm that expedites scientific discoveries and makes them broadly accessible.
Past and ongoing projects are provided below (note this is only periodically updated).
Unifying Infinite-Dimensional Optimization
Ongoing Project
Infinite-dimensional optimization encompassess a wide breadth of challenging decision-making problems that entail decisions indexed over continuous domains (e.g., uncertainty and/or space-time). Applications include process control, climate modeling, bio-manufacturing, and conditioning of large indoor spaces. We have developed a unifying modeling abstraction for these problems that accelerates scientific descovery and accelerates their solution on CPU and GPUs.
Machine Learning Surrogate Models
Ongoing Project
Machine learning models such as neural networks have proven highly versatile in a wide variety of engineering applications. We are developing neural network models that guarantee constraint statisfaction, enabling us to embed domain knowledge via constraints that increase model robustness, reduce overfitting, and reduce reliance on large datasets. Moreover, we are developing software to efficiently train these models on GPU and embed them in large-scale optimization models that can be solved with rapid 2nd-order GPU solvers.
PDE-Constrained Optimization of Conditioned Spaces
Ongoing Project
Balancing occupant comfort against energy use, greenhouse gases emissions, costs, and environmental impact is not trivial in large indoor conditioned spaces. Ice rinks in particular are challenging since they exhibit significant variability in temperature and humidity, time-varied occupancy, weather-driven disturbances, and volatile electricity pricing. we are developing an optimization framework that jointly considers energy costs and occupant comfort under realistic operational constraints while remaining computationally tractable.
Center for Innovative Recycling and Circular Economy (CIRCLE)
Ongoing Project
This project is part of an international global centre with over 40 researchers from 18 academic institutions and various industry partners. The goal is to develop sustainable bioprocesses to reduce pollution and contribute to a circular economy by converting different waste streams into valuable products. Our contribution will be focused on making software tools to aid in optimized process design while incorporating environmental burden, and developing better methodologies for creating robust models of complex biological processes.
Data Framework for High-Dimensional Compartment Fire Data
Ongoing Project
Modern airtight residential construction has fundamentally changed how house fires behave, producing more toxic and unpredictable under-ventilated fires that our codes and common practices weren't designed for. We developed a data-driven framework to analyze high-dimensional gas sensor data from full-scale compartment fire experiments, enabling us to systematically identify stages of under-ventilated compartment fires, identify which chemical species most strongly correlate with fire intensity, detect faulty sensor measurements, and reliably detect regime transitions using deployable sensors.
Random Field Optimization
Past Project
Random field theory provides a powerful mathematical framework for characterizing uncertainty over space-time. This enables us to better capture real-world infinite-dimensiional systems that often exhibit random phenomena. We have established and are actively developing a framework which we call random field optimization that incorporates random field uncertainty into decision-making problems.
Computer Vision for Process Control
Past Project
Computer vision aided process control architecture is increasingly used in industy, but a significant challenge to widespread deployment is accurately monitoring sensor performance in real-time. We propose the SAFE-OCC framework to tackle this challenge and deploy these sensors with confidence.
Resilient Energy Infrastructure
Past Project
Rare, high-impact events (e.g., extreme weather, international conflicts) can inflict severe disruption/damage to energy infrastructure. Mathematical strategies are needed to characterize the effect of rare events on energy infrastructure such that we can economically hedge against catastrophic failures. My research has focused on developing frameworks for assessing and designing the flexibiity and reliability of critical infrastructure networks (e.g., power grids).
Optimal Rare-Earth Element Recovery Wastewater Networks
Past Project
Industrial wastewaters (e.g., produced water from oil/gas production and acid mine runoff) are often prohibitively expense to fully treat to mitigate environmental impact and reuse. However, these wastewater streams can contain appreciable amounts of valueable rare-earth elements and critical materials (REE-CMs). In collaboration with NETL, we helped develop a decision-making framework to explore opportunities for wastewater treatment and reuse when REE-CM recovery is taken into account.
Disease Control and Analysis
Past Project
Recent disease outbreaks (e.g., Covid-19) have highlighted the need for enhanced analysis and modeling of disease spread to better inform public policy on effective mitigation measures. My research has developed scalable methods to effectively analyze disease behavior with models that can be used to optimally inform disease control measures.
Unmanned Aerial Vehicle Driven Infrastructure Monitoring
Past Project
Aging infrastructure accross the U.S. and beyond needs to be regularly inspected to prevent catastophic failures like that of the flood levees in New Orleans in 2005; however, with limited manpower and resources regular inspections are simply not possible. We addressed this problem using unmanned aerial vehicles that can autonomously inspect infrastructure with optimal flight paths to maximize information gathering and automatically detect changes that require human intervention.