What I have built, in full detail.

DS 2500 capstone, four-person team

Do regulatory penalties change bank behavior?

• Collected roughly 35,000 consumer complaints through the CFPB API, handling paginated JSON and restructuring nested payloads into analysis-ready tables.
• Compared complaint patterns at Wells Fargo and TD Bank around major enforcement events.
• Built time-series visualizations and institution-level breakdowns.

My first complete end-to-end data science pipeline.

Solo build

A mobile-first poker and blackjack decision trainer. The interesting part is the simulation, not the card game.

• Monte Carlo equity simulator running in cancellable Web Workers, reporting 95% confidence intervals across Fast, Balanced, and Precise modes, with caching to avoid redundant computation.
• Rule-aware blackjack basic strategy across deck counts, H17/S17, double-after-split, and late surrender.
• Validated dollar-stakes table engine with legal-action validation, side pots, and street advancement for 2 to 10 players.
• Account-backed training history and profit/loss ledger on Supabase Postgres with row-level security. Guest mode works fully offline.
• Deterministic unit tests with coverage gates, verified on every push by GitHub Actions.

Solo build

A privacy-first Instagram data export analyzer. The export is parsed locally in the browser and the raw file never leaves the device.

• Local ZIP and JSON parsing with JSZip, handling export formats that vary by region and app version.
• Unified social graph connecting follows, DMs, and interactions, with scoring for relationship strength, cleanup priority, and privacy exposure.
• Day-by-hour activity heatmaps, monthly timelines, and engagement breakdowns rendered with Recharts.
• Optional Supabase cloud save that stores computed insights only, never raw messages or media.
• Fail-closed Presentation Mode that blocks identity, DM, and search views during screen sharing.

No upload needed. Ships with synthetic demo data.

DATA Club bootcamp, four-person team

Ten seasons of NFL salary allocation (2013 to 2022) modeled against team results, with an interactive dashboard.

• Assembled a decade of salary cap allocation and season outcomes into a single modeling dataset.
• Modeled spending by position group against team results.
• Built an interactive dashboard for exploring the relationship season by season.

Solo build

The site you are reading, built to be the one link on my resume.

• Publishes every course report, presentation, model, and poster as a linkable document.
• Server-rendered on the App Router with a light and dark theme, indexed through Google Search Console.