Motherlode
A data pipeline that keeps a Canadian mining consultancy's deal research up to date on its own. Nobody has to run it.
Challenge
Their analysts tracked mining M&A by reading news sites by hand. It was slow, and deals slipped through.
Solution
A pipeline that scrapes the sources with Playwright, classifies and extracts each article with an LLM, and loads clean records into PostgreSQL. GitHub Actions runs it on a schedule and commits the results back to the repo. No servers to babysit.