The product: a structured phytochemical dataset derived from the USDA Dr. Duke's ethnobotanical database, denormalized and enriched with quantitative signals from PubMed, ClinicalTrials.gov, ChEMBL, and USPTO PatentsView. Stack: PostgreSQL 15, Python 3.12 (async httpx pipelines), Rust/Actix-Web, Cloudflare Workers/R2.

Three weeks ago I had no idea what a Parquet file was. I stumbled onto a data gap in natural product research and decided to close it.

Interested in the gap between public-domain research databases and production-ready data products — and in what's possible when AI handles the implementation.