Reddit
Hiring.Cafe scraped 5.9 million jobs straight from company career pages by feeding raw domains to an LLM
The builder behind Hiring.Cafe posted an update (70 comments) saying the site now holds 5.9 million jobs including 220,000 remote roles, all pulled from companies' own career pages rather than LinkedIn or Indeed, which he argues are saturated with ghost postings and offshore agency reposts. The method sidesteps the usual per-site scraper problem: dump raw domain names at the model and have it return normalized JSON with salary, years of experience, and the rest, regardless of each page's structure. He open-sourced the extraction prompt as a GitHub gist. That prompt is the reusable artifact here for anyone doing heterogeneous web-to-schema extraction, job board or not.
↳ Follow the thread