Personal Projects

• Q-Learning AI for Tic-Tac-Toe

I became interested in learning about Q-learning methods after watching several Youtube videos, and successfully built my own AI to self-play the game Tic-Tac-Toe. After several million training rounds, the AI always tied or won against itself.

Github

• Q-Learning AI for SushiGo

After my Tic-Tac-Toe project, I attempted to create another agent to play SushiGo, which is a nondeterministic multiplayer card  game, similar to Poker. Currently the AI slowly learns and reached a peak winrate of 70% against players who play randomly, but failed to find a winning strategy against competent opponents.

Github

• Melting Point prediction from Coulomb Matrices

In the summer of 2024, I worked with UTORII graduate student Nick Furth to investigate which machine learning model (Dense neural network, Convolutional neural network, or XGBoost) had the lowest average error when predicting melting points using Coulomb Matrices as inputs. After I presented our findings at the end of the summer, I have continued to tweak the models to lower the error. We have developed a model with MAE of 24 and are currently working on publishing our work. Github coming soon!

• Kattis Programming Problems

I am currently ranked #3 in Tennessee and #1642 / 35571 globally on the programming problem website Kattis, where I have completed around 250 problems. My profile

Schoolwork Projects