(4hr) Bridging Decision-Making at the Microscopic and Macroscopic Levels using Heterogenous Modeling and Optimization

In my research career to date, I have worked in several key areas of chemical process systems engineering. At the University of Texas at Austin (UT), my PhD work at the McKetta Department of Chemical Engineering involved process design optimization, demand response scheduling (i.e., flexible operations), and decision-making under uncertainty. I was then awarded fellowships by the EPSRC and Imperial College to join the Department of Computing at Imperial College London as an independent Research Fellow. These awards allowed me to broaden my perspective by transitioning to a new field in starting my independent research career. At Imperial, my work focuses on data-driven optimization, optimization software, and mixed-integer programming. My projects have emphasized real-world applications, and many involved extensive industrial interactions/feedback via the Texas-Wisconsin-California Control Consortium (UT) and the Sargent Centre for Process Systems Engineering (Imperial).

Research Interests:

A common theme of my research to-date is the development of adjustable optimization models: in other words, model size is tuned to balance quality and tractability. However, in multiscale optimization problems that bridge decision-making at the laboratory and process systems levels, a single model may not suit all considered scales. For instance, finding the optimal system configuration for a new bench-scale development may require accurately modeling microscopic chemical engineering phenomena, but remaining scalable to macroscopic, system-wide optimization. Therefore, I plan to explore strategies that creatively deploy multiple heterogenous models (i.e., different instances of an adjustable model) in multiscale optimization, rather than just selecting one. Below, I describe three general research topics concerning this theme.

1. I) Optimization Tools for Materials and Reaction Engineering.

My research here seeks design-of-experiments (e.g., Bayesian optimization) techniques that bridge chemical engineering experiments with system-level analyses. My group will design optimization methods that link standard measurable targets and process-level performance. Furthermore, the methods will integrate with two important properties of process systems: the availability of historical data and the presence of non-trivial constraints. This interdisciplinary work will establish connections among process systems engineering, machine learning, and optimization.

1. II) Expanding Flexible Operations in Chemical Processes.

Modern chemical processes often must deviate from traditional steady-state operation paradigms. Here, optimization can reveal the flexibility potential of new technologies, as well as guide future developments. While such optimization procedures usually rely on a model that balances accuracy and tractability, my group will investigate deploying heterogenous models to improve accuracy and/or computational performance. This may involve quantifying accuracy/optimality, formulating linking/complicating constraints between heterogenous models, and tailored optimization schemes.

III) Computational Techniques for Supply Chain Optimization.

My research here seeks to systematically deploy heterogenous relaxations and restrictions during global optimization to accelerate convergence. My group will initially focus on features common to chemical supply chain formulations: binary variables (used to model on/off decisions) and bilinear terms (present in mixing equations). This research will have strong algorithmic and software components, e.g., how to effectively add/remove auxiliary variables, initialize their values, and tighten their bounds.

Teaching Interests

A chemical engineering education prepares students for a lifetime of critical thinking, problem-solving, and discovery. Therefore, I believe education comprises not only engineering skills and knowledge, but also mentorship in asking new questions and finding new applications. Given the above, I aim to integrate my teaching with real-world connections and challenges from my personal (and secondhand) experience. I continually monitor teaching methods (e.g., activities, examples, laboratories) in my area, in order to trial and implement tools that suit my teaching style and student interests.

My teaching experience at UT involved two semesters as teaching assistant for Process Dynamics and Control, where I ran recitation sessions, graded exams, and gave office hours. After taking a Teaching Practicum course at UT, I was teaching assistant for a new Product Design and Commercialization course, where I helped design the syllabus and course materials and coordinate the semester-long projects. At Imperial, I developed my mentoring skills by supervising three computing MSc theses and co-supervising a PhD student.

With extensive experience in both Chemical Process Systems Engineering and Computational Optimization, I am well-suited to teach a wide range of courses. I am particularly interested in courses related to computational tools, e.g., numerical methods, process dynamics and control, optimization, applied statistics, process design. Furthermore, I am excited to develop elective courses in process modeling and optimization, as well as data science and advanced optimization; I believe these topics are increasingly important in chemical engineering education.