Projects & Research

Implemented Meta's Prophet across six crime categories and 90+ Toledo census tracts on 500K+ records (~85% accuracy in project evaluation), capturing seasonality and holiday effects

Paired forecasts with automated ETL and an interactive Shiny surface for practitioner exploration

Presented neighborhood-level forecasting methods and policy implications at ACJS 2025

Built a reproducible training/ops pipeline with DVC and GitHub Actions (framed with Docker / MLflow / Airflow patterns) for clinical imaging models

Combined topological data analysis with deep learning for tumor detection; ensemble approach reached ~92% classification accuracy in reported evaluation

Added monitoring and retraining hooks aimed at production-minded model lifecycle management

Interactive environment for exploring data science concepts through hands-on Clojure notebooks

Implementation using Scicloj/Clay and Quarto for web-based notebooks

Docker containerization for reproducible environments

Comprehensive coverage of statistical and machine learning topics

Implemented numerically stable algorithms for statistical computations

Developed eigendecomposition methods and PCA implementation

Comparative analysis against scikit-learn implementations

Utilized Neanderthal, tablecloth, and scicloj.clay libraries

Implemented convolutional autoencoders for image denoising

Compared dimensionality reduction techniques on Fashion-MNIST

Built with PyTorch, scikit-learn, and Clojure/Clay

Created promotional video highlighting university features and programs