Data Engineer at Travelers Insurance and MS Analytics student at Georgia Tech, based in St. Paul, MN. I build governance pipelines and cloud automations on Databricks and AWS.
- Twin Cities Living Quality Map: an interactive map scoring every St. Paul and Minneapolis district from 0–100 on safety, opportunity, amenities, transportation and economics, built from a Python pipeline over city, OpenStreetMap, Census, MnDOT and CDC data (live site)
- Gopher X Metro: one live map of every University of Minnesota campus bus alongside Metro Transit buses and light rail, built in 2024 with a team from Social Coding at UMN. In 2026 I revived it after the original hosting lapsed, rebuilding it to run free with no API keys (live site). Companion project Will My Bus Be Late? logs Metro Transit lateness every 15 minutes and retrains LightGBM models daily to flag departures that usually run late (live site)
- Will My Flight Be Delayed?: enter a US flight number and date to get the chance it arrives 15+ minutes late, from a LightGBM model trained on 7M flights using route history, FAA aircraft data, the plane's inbound flight and hourly weather. On two months of flights it never saw, it picked out delayed flights 73% of the time, versus 62% using past on-time records alone (live site)
- Minneapolis Home Assessment Check: compares a home's assessed value with time-adjusted sales of similar nearby homes, plus an independent gradient-boosted price estimate with calibrated (conformal) ranges. Its typical error in the 2026 backtest was 9.4%, versus 12.0% for the city's values (live site)
- Movie Recommender: rate a few movies and get a ranked watch list. I benchmarked eight recommenders on 33.8M MovieLens ratings, and the winner, EASE, ranks relevant movies about 7x better than a standard matrix-factorization baseline. It runs entirely in the browser and retrains monthly when MovieLens updates. It grew out of my DSCI-4093 capstone (live site)
- My Social Media, as Data: a data story from 14 years of my Instagram, Reddit, YouTube, Snapchat and Twitter exports plus an iPhone backup (iMessage and Apple Health), publishing only aggregates. It shows how recommendation algorithms took over my feeds (my top 10 accounts went from most of my likes to under 10%) and contrasts that with texting, where my top 10 conversations get 89% of my messages. Uses sentence embeddings and RoBERTa sentiment (live site)
- Creekside: a couch co-op pixel puzzle-platformer for two players on PS5 controllers, with ten chapters, procedural pixel art and synthesized chiptune audio (play it)