Every weeknight rfp.rrecktek.com pulls the day's federal grant postings into a single Postgres table. This is the story of one organization's path through it — from 122,230 opportunities to the 22 grants actually worth writing.
Forty-three scanned medical sheets, understood by a 3-billion-parameter model running on an on-premise 4080 GPU, turned into 208 searchable vectors and 45 tracked lab values — in eleven minutes, with nothing sent to the cloud. The pipeline, the timings, and the numbers.
The complete DOJ Epstein release is 1,378,652 documents. Instead of reading headlines about them, we indexed, embedded, and machine-read the entire corpus — full-text search, semantic vectors, and computer vision — measuring 46 topics, 48 named individuals, and every image, to establish what is actually in the files and what is not.