Projects & Research

Parkinson's Disease Classification App - AI4ALL Ignite

Python, Streamlit, Scikit-Learn, Pandas, Responsible AI

Developed as an AI4ALL Ignite Fellow. Led a 7-person team to train decision-tree-based models achieving 90% accuracy in classifying Parkinson's Disease. Deployed via Streamlit with embedded fairness audits and responsible AI checkpoints. Selected as the recipient of the Innovation Award at the AI4ALL Student Symposium.

Heart of the Home - Capstone

React, MySQL, TypeScript

Completed senior capstone project building an accessible, mobile-friendly application to streamline and strengthen Habitat for Humanity Greater Peoria Area’s connection with admins and homeowners. Led frontend development, core real-time dashboards, resource/event modules, and integrated the MySQL backend.

WiDS Datathon 2026: Wildfire Spread Prediction

Python, Scikit-Learn, NumPy, Matplotlib, Pandas

Engineered physics-informed features including Wavefront ETA (combining radial growth and centroid movement) to capture non-linear fire dynamics. Designed a 50/50 blend of a Random Survival Forest and a 4-model Gradient Boosting ensemble, achieving a localized Brier Score of 0.003.

Entrepreneur Compass Tool - Cambio Labs

Python, SciKit Learn, Pandas, Numpy

Developed during the Break Through Tech AI Program. Conducted EDA and engineered synthetic data samples to handle sparsity, optimizing Random Forest and KNN models to hit 80% recommendation accuracy for low-income entrepreneurs.

HoodWatch - Neighborhood Safety App

Figma, HTML, CSS, React, Expo SDK, MongoDB, TypeScript

Created to report local safety issues in real-time. As Frontend Lead and Researcher, designed and implemented the cross-platform mobile UI using Expo SDK and React Native, setting up navigation, auth, and secure data handling.

Twitter Sentiment Analysis

Python, Jupyter Notebook, NLTK, Scikit-learn, Pandas

Optimized Random Forest and Naive Bayes models to predict sentiment and reshare likelihood. Built word clouds using NLTK; the Naive Bayes binary model achieved a 93.31% accuracy with oversampling.