Portfolio
Software, data, and AI projects—with a closer look at what I contributed.
01 / Applied AI · Team project
An AI booking assistant and CRM prototype developed for Everything Dough, a small business in Stamford, Connecticut.
My contribution. Led team strategy, prompt engineering, and the live presentation. Lexi handles booking inquiries; Crust brings lead and pipeline information into a dashboard.
A hackathon prototype demonstrated live to nearly 50 attendees at AI for Impact.
02 / Machine learning · Capstone
An end-to-end audio pipeline with a local app that returns a track’s top three predicted genres.
My contribution. Led the team and worked across ticket planning, PR reviews, CI/CD, database design, preprocessing, and model development. Built with Python, PostgreSQL, PyTorch, FastAPI, and JavaScript.
Trained a three-block CNN on GTZAN audio, reaching 30% test accuracy across 10 genres. The repository documents generalization limits and next steps; the demo runs locally.
03 / Data analysis · Classification
Exploring fraud signals in a dataset of more than six million bank transactions.
My contribution. Performed exploratory analysis, engineered balance-related features, and built and evaluated a classifier. Investigated how threshold tuning changes the precision–recall trade-off.
Reported ROC AUC of 0.9972. Lowering the decision threshold from 0.50 to 0.30 raised recall to 77%, while precision moved from 96% to 94%.
Exploring song recommendations through lyrical features, dimensionality reduction, and clustering.
Comparing regression approaches and documenting the limits of weather-based predictors.