Designed and coded a 25,000 parameter physics-informed neural network to decompose the capacity fade of lithium-ion batteries into specific electrochemical processes. Employed a two-step training process, during which penalization from physics-fidelity and regularization terms is gradually added to the loss function in order to guide the model towards physically meaningful solutions. The algorithm achieved an MAE of less than 1%, demonstrating its effectiveness

Terra NYC STEM Fair 2026 - 1st Place, Engineering

Undergraduate group research (REU) to motivate a proprietary, generalized shell model of turbulence. As part of the group, I wrote several programs in Julia that compare the energy and flux spectra of our model to those of more classical models from earlier literature (ie the Sabra model) in order to validate the dynamics and long-term behavior of our model

Presented at JMM 2026 in Washington DC

REU Advisors Prof. V. Martinez, Hunter College & Prof. C. Victor, Texas A&M

Independent neuroscience research to build proprietary stochastic gradient descent algorithm that leverages L-BFGS-B and Adam optimizer techniques to estimate the parameters of the Fitzhugh-Nagumo system of ODEs. Model achieved an MSE of less than 10^-9, demonstrating its reliability in solving the inverse problem

Terra NYC STEM Fair 2025 - 2nd Place, Computational Biology

My coding partner and I wrote a solution to a 3500-rated Codeforce problem (the highest-level difficulty), requiring strong proficiency in dynamic programming (and guidance from our advisor & TAs); presented as our capstone project for the program