Projects

Physics-Informed Neural Networks (PINNs)

Tests whether building physics into a neural network helps predict how long 3D-printed metal parts last before cracking. Across 8 datasets (30 to 10,000+ samples), it did not consistently beat standard models. Data quality mattered more.
Abstract submitted to Fatigue Design 2027.

Tools: Python, PyTorch, scikit-learn, Monte Carlo simulation, AI, machine learning

See Project

NLP & LSTM Language Model

Analyzes the words and emotional tone of 250+ songs, then trains an AI to write new lyrics in a similar style by predicting the next word.

Tools: Python, NLTK, LSTM neural network, Tableau, AI, machine learning

See Project

Predictive Modeling of California’s Energy Systems

Forecasts California’s energy production and consumption from 20 years of state data. I compared 7 models, and the best one matched the data almost exactly, with an RMSE of 4.9e-10.

Tools: Python, Google Colab, Tableau, regression, machine learning

See Project

See Project

Student Rental: Use Case Study Report

Designs a database for a student housing rental service to keep information organized and easy to search. Built and tested in MySQL, MongoDB, and Python.

Tools: MySQL, MongoDB, Python

Photovoltaics: The Science, History, and Future

Explains how solar panels work and weighs the strengths and limits of solar energy, from the physics to the economics. I presented it to 40 students as a guest lecturer.

Tools: research, circuit diagrams, presentations

See Project