I was a course instructor for Calculus III (MAT235), Calculus I (MAT186), and Differential Equations (MAT234). Some were taught in an online setting during the COVID-19 pandemic, and others were taught either hybrid or fully in-person. I worked alongside teams of instructors to design course material and assessments, manage teaching assistants to run tutorials, invigilate exams, and grade assignments, as well as perform other administrative duties.

Inspired by advancements in the field of Scientific Machine Learning, I designed a deep reinforcement learning algorithm that improves upon the classical Monte Carlo schemes for the Feynman-Kac formula. Written in python and using the pytorch machine learning library, the self-supervised deep learning algorithm trains a neural network that solves initial boundary value problems of quasilinear parabolic partial differential equations in irregularly shaped domains. The algorithm is mesh-free and generalizes well in higher dimensions. Furthermore, I established convergence results and error bounds from statistical analysis of the hyperparameters. On a modest workstation, 2 digits of accuracy were achieved in under 2 hours.