A full-stack social web app I designed, built, and deployed independently: star any webpage, post to a per-page forum, and join a per-page live chat room. Backed by a production PostgreSQL database with Redis and Celery for async tasks, on a DigitalOcean Linux server, paired with a Chrome extension published to the Web Store.

An end-to-end machine-learning pipeline (for a Kaggle competition) predicting problematic smartphone and internet use from survey and behavioral data: cleaning and quality checks, feature engineering, a logistic-regression baseline and a tuned LightGBM model, evaluated with ROC-AUC.

A personal website and blog built in Next.js and React, hosted on Vercel with continuous deployment from GitHub. Home/bio, CV, and writing.

A Django web application publishing experimental "hyperfiction": creative writing that incorporates internal and external hyperlinks, popups, sound, and images.

Used similarity-based networks to analyze existing music genres and construct new ones: built a network from song audio features, then applied spectral clustering and modularity maximization, validating communities against ground-truth genre labels.

Reproduced ProPublica's analysis of racial bias in the COMPAS recidivism algorithm, then trained Decision Tree and SVM classifiers to compare fairness across multiple metrics (false-positive/false-negative rates, cost-based formulations), showing how competing fairness definitions yield conflicting conclusions.

A Python application that streams tweets and scores their sentiment against a crowdsourced lexicon to visualize real-time, state-by-state emotional conditions across the United States, rendered onto a US map.

A graphical simulation of disease spread across a population, with customizable transmission rate and range, incubation/illness/convalescence periods, death rate, and virus mutation rate.

Generates and saves a word graph from a corpus of text files, storing each word with its frequency and co-occurrence counts: a foundation for NLP and network analysis of language.

Scans text documents to flag misspelled words with suggested fixes and recommends synonyms based on each word's context.

Generates random 3D objects based on a chosen number of vertices and lets the user inspect them interactively through rotation: a graphics and computational-geometry exercise.