JevMade hello@JevMade.com
← Back to guides

JevMade field notes / Application walkthrough

Run a local YC company search pipeline with Jev

An Apple Silicon application that searches a committed YC company index with text, logo, OCR, and tag signals, then uses Jev to rank a bounded finalist set.

Original by Aayan-DEVRetrievalGitHub README and search modulesSource reviewed

Before you dive in

What you’ll find in the original

  1. Run the committed index on an Apple Silicon Mac; rebuilding the company data is optional, but every query depends on the bundled Core ML embedding binary.
  2. Combine local meaning, logo, OCR, and YC-tag signals to narrow 6,241 companies to at most 320 finalists.
  3. Rank only that reduced set with Jev, and retain the documented local fallback when no TypeSafe key is configured.
Worth knowing

Requires Node 20+ and an Apple Silicon Mac; a TypeSafe key enables Jev ranking, while keyless runs use a degraded local scorer. Search quality, timing, and cost are project-reported and were not independently benchmarked.

More guides →See the experiments →