(3ib) Uncertainty-Aware Design: Bayesian Modeling in Molecular-Level Chemical Engineering

My research integrates probability theory and machine learning with fundamental problems in chemical engineering, with a particular emphasis on uncertainty-aware modeling. During my PhD and postdoctoral work, I have focused on Bayesian methods such as Gaussian process regression, classification, and Bayesian optimization to address challenges in molecular simulation and experimental data interpretation. These methods provide not only accurate predictions, but also principled estimates of uncertainty, enabling smarter decisions about experiment design, data collection, and model improvement.

A central theme of my work is the development of physically grounded and data-driven models for molecular interactions. Some applications of my research include:

• Uncertainty-aware machine learning potentials based on Gaussian processes
• Analysis of experimental scattering data using non-stationary Gaussian processes
• Bayesian force field optimization for aqueous ionic systems
• Theoretical study of electron polarization for next-generation interatomic potentials.

My current research applies these approaches to biomolecular systems involving complex macromolecule-ion interactions, with applications to drug design, materials discovery, and understanding cellular-scale processes. I am also exploring “simulation-free” prediction of molecular dynamics using Fokker-Planck equations in ongoing collaborative work on non-equilibrium statistical mechanics.

As a faculty member, I aim to lead a research group that applies state-of-the-art probabilistic machine learning and decision-theoretic tools to contemporary chemical engineering problems. Beyond force field development and simulation, my group will develop novel strategies tailored to specific challenges in molecular modeling, drug discovery, and materials design. With expertise spanning chemistry, physics, and data science, our group will also be well-positioned to collaborate across disciplines, helping the department make better use of data and pushing forward the integration of uncertainty quantification into engineering design and discovery.

Teaching Interests:

Teaching is one of my greatest passions. I am especially interested in teaching the subjects thermodynamics and statistical mechanics, although I also have experience as a teaching assistant for courses on fluid mechanics, heat and mass transfer, and machine learning. Furthermore, I have already written and taught my own upper undergraduate / graduate course on statistical mechanics and molecular simulation at the University of Utah (link to course notes: https://bshanks.netlify.app/course/mol-sim/). I prefer a discovery-based approach to teaching in which we follow a (nearly) historical path through the material, starting with early and easy concepts and gradually posing questions that we will later resolve with new mathematical and conceptual tools. I like to emphasize “learning how to learn” by tailoring lectures, homeworks, and exams to require critical thinking about the problem rather than mechanically repeating classic problems. I also like to emphasize concepts over memorization, with the aim of leaving students with a long-lasting understanding of the core ideas of a subject rather than forcing them to memorize information that they will inevitably forget. Finally, I like to provide a diverse array of study materials, including: in-class lectures, lecture notes, videos, recommended reading materials, and interactive python codes, to ensure that the majority of students can find an effective way to approach the material. The primary objective of my courses is to leave students with enough knowledge to pursue more advanced study on the subject on their own, aiming to create independent thinkers that know how to learn and teach themselves.