Current
askdata.app — natural language for structured data
askdata.app is an LLM-powered interface that lets users explore datasets by asking questions in plain English. You connect a dataset, ask a question, and get back an answer, a chart, and the query that produced it – so you can check the work.
The interesting hard parts are schema understanding, hallucination control, and evaluation. Status: working prototype. Read the build notes →
Selected past work
Bernoulli One — continuous patient monitoring
Built a custom physiological indicator that detected severe events 12 seconds earlier than the industry standard at equal accuracy – enough lead time to change clinical action. Time-series and data mining on continuous patient monitoring data. Details.
BUILD — North Star for a national nonprofit
Helped a national youth-serving nonprofit shift their North Star from program count to student outcomes, then built the automated reporting connecting it to their data sources. Write-up.
Machine Learning for Good — open-source ML curriculum
Contributed to Delta Analytics’ open-source ML course for nonprofit data practitioners. GitHub.
Handy — R package for data cleaning
An R package for common data cleaning and exploration tasks – name cleaning, variable summaries, human-readable plot scales, Excel output.
Older explorations
- NCAA March Madness Elo Rankings – Kaggle analysis predicting tournament outcomes from historical performance
- Foosball Elo Analysis – rating system for office foosball, built as a Shiny app
- Kiva Loan Analysis – interactive visualization of microloan data