Featured Projects
Self-developed, deployed builds.
A live web app that helps instructors form balanced student teams from a quick survey. It grew out of my Birds of a Feather research — four matching models (Hungarian assignment as the optimized default), a six-metric team-quality score, and a privacy-first design that processes data in memory only.
A normalized SQL database of 1.35M consumer loans, with scikit-learn models that flag loan default at 0.716 ROC-AUC and an interactive Tableau dashboard of the risk segments — including a borrower-only model that edges out the lender's own credit grades.
Case Studies
Analytical studies with full writeups.
The published research behind Teamora: an ML pipeline that forms fairer student project teams with clustering and Hungarian assignment, audited for the bias hidden in self-reported skills and GPA.
A machine-learning approach to the NCAA tournament: feature engineering on regular-season team stats, model comparison, and evaluation on real bracket outcomes.
An exploratory analysis of NFL home-field advantage from 2000–2023 — measuring the size of the edge and how it has shifted across more than two decades.