BlueRes Search
Résumés and job posts are public data — so why shouldn't the matching process be transparent too? BlueRes Search is a proof-of-concept that uses semantic embeddings to search and match job posts and résumés on the AT Protocol, without relying on a black-box AI.
Both document types are broken into structured fields, embedded into a shared vector space, and retrieved or compared by similarity. The goal is to show that job-to-candidate matching can be done in a way that is open, inspectable, and clearly articulated — no magic required.
Search
Query job posts or résumés by meaning, not just keywords. Choose between nearest-neighbour, top-K chunk, and multi-query modes to control how results are ranked and deduplicated.
Match
Compare two documents — or free-form text — by embedding distance. For each section in a reference document, find the closest match in the other and average the scores.
How it works
- 1. Structure. Raw job posts and résumés are parsed into typed JSON schemas — job posts using a custom schema, résumés using jsonresume.
- 2. Embed. Configurable extractors pull the text fields that matter, each chunk is passed to an embedding model, and the vectors are stored alongside the source document.
- 3. Search or match. A query or reference document is embedded the same way. Search returns the closest documents; matching scores each section pair and averages the result.