Student-built solutions.
Real nonprofit impact.

An interactive data visualization tool for Strong Towns Madison, built around the City of Madison's tax parcel data. The project pairs a map view that colors parcels by net taxes, taxes per square foot, or land use, with a separate statistics page offering richer filtering and several chart types, both sharing the same filter system.

The frontend is built with Svelte and deployed on Cloudflare Workers, using MapLibre GL JS for the map and Layercake for charts. Instead of running a live backend, GitHub Actions pre-process the source .parquet data into static files hosted on Cloudflare R2: a .pmtiles file (generated with Tippecanoe) powers tiered map reads, and a JSON dump backs the stats page. The shared frontend with the Tax Explorer team allowed both projects to ship a cohesive experience.

A property tax exploration tool for Madison residents, built in response to a recent referendum that significantly expanded property-tax funding for Madison schools. Users can search any Madison address to view multi-year assessment history, effective tax rates, and a breakdown across City, School District, County, and MATC, plus land-efficiency metrics like land value per square foot and taxes per square foot.

The frontend is SvelteKit (TypeScript) on Cloudflare Workers with LayerCake for visualization. The backend is a Rust/Axum REST API on Google Cloud Run, backed by DuckDB for fast analytical queries against Madison's property tax roll data. DuckDB turned out to be a great choice for multi-year aggregations without the overhead of a heavyweight DB server.

Badger Prairie Needs Network is the third-largest food pantry in Dane County. To fund its work, BPNN runs holiday flag fundraisers that deliver flags to roughly 400 addresses across ~24 volunteer driver teams. This project automates the otherwise manual task of carving those addresses into balanced, efficient driver routes.

Volunteers upload a spreadsheet of addresses; a Python service geocodes them and clusters via K-Means or DBSCAN (with haversine distance) into one cluster per driver. A distance matrix is then built per cluster via the OSRM API and fed into Google OR-Tools to solve each cluster's traveling-salesman route. Routes are returned to the user on an interactive map, can be reassigned between drivers, named, saved to Supabase, and exported as JSON. The frontend is React + TypeScript + Vite, deployed via Vercel.

Sustain Dane is a Dane County sustainability nonprofit whose Efficiency Navigator Program helps residents make their homes more energy efficient. This project is an interactive map for homeowners and renters to see, at the neighborhood level, projected dollar savings, kWh saved, vehicle miles avoided, and CO₂ emissions savings from previous Sustain Dane renovations near them.

Users can share their location or search an address; the app routes to the nearest known intersection node and surfaces donut visualizations and plain-language explanations in a sidebar. To keep geocoding cheap, lookups use a three-step cascade: a precomputed JSON of intersection coordinates, a cached browser session, then a live Photon geocode (with regex normalization for typos and abbreviation variants) as a fallback. Built with React + Vite, Leaflet, and Tailwind CSS.