Internal web application built for a gym's staff and trainers to log classes and hours, and for owners to track equipment status via member-facing QR codes that report issues in real time. (Client name, branding, personal information, equipment data, and QR code contents have been redacted/replaced for this portfolio.)

Simulates core operating system scheduling concepts applied to a healthcare setting: FCFS, Round Robin, and Priority scheduling algorithms, plus ICU bed allocation logic.

Real-time market sentiment dashboard that aggregates Reddit, Google News, and Stocktwits data, scores it with VADER sentiment analysis and LDA topic modeling, and overlays sentiment against stock price history.

End-to-end subscriber churn analysis for a streaming platform: monthly churn by subscription tier, revenue at risk, cohort survival curves at 3/6/12 months, and customer lifetime value by acquisition channel, presented in an interactive Tableau dashboard.

Upload a PDF and get a quiz that adapts to your weak spots: topics are auto-discovered via K-means clustering on sentence embeddings, and per-topic mastery is tracked with an exponential moving average. Runs against a local LLM backend (Ollama or Apple MLX).

A custom programming language and interpreter built from scratch, supporting variable declarations, control structures (if/else, while loops), functions, and error handling.

Flask app that flags likely misinformation in a block of text using a trained classifier.

Encrypted client-server messaging app: RSA key exchange for secure session setup, AES message encryption, and a lightweight anomaly detector that flags suspicious messages (injection patterns, abnormal text entropy) in real time.

Log analysis and alerting pipeline: parses and sorts raw log files, flags anomalies against expected system-event patterns, visualizes anomaly trends, and includes a small GUI for browsing results.

Mobile weather-guessing game: players guess a city's current temperature at increasing difficulty tiers, build daily streaks, and compete on a leaderboard, backed by a live weather API.

AI-powered legal document analyzer that autonomously reads, interprets, and summarizes legal documents, extracts key clauses, detects risks, and generates brief reports — built to save time for legal professionals on tedious document review.

AI-powered car-finding system built from three specialized agents: a scraping agent that pulls listings from Craigslist, AutoTrader, and Cars.com, a filtering agent that scores listings against user preferences with weighted ranking, and an agent that drafts negotiation emails.

Analyzes hospital patient data to identify patterns in readmission rates, with an interactive dashboard surfacing insights for healthcare improvement.

Predicts traffic congestion across Philadelphia and surfaces actionable insights for optimal routing and urban planning.

Early-stage project to classify clinical text into ICD-10 diagnostic codes: text preprocessing pipeline and initial exploratory analysis of a sample clinical dataset. Model training in progress.

In-progress exploratory analysis of a used-car listings dataset: cleaning and standardizing price, mileage, and fuel-type fields as a first step toward market pricing insights.

AI-assisted car specification lookup tool: falls back to OpenAI to generate realistic multi-year, multi-trim vehicle specs (power, transmission, drivetrain) when the CarQuery/NHTSA APIs return no data, with in-memory caching to avoid redundant calls.

Small-business bookkeeping web app: dashboard with cash-flow charts, expense and deposit tracking, receipt management, vendor records, and multi-org support behind Supabase auth.