Candidate Sourcing And Matching

Candidate uploads a multi-column PDF resume with a research-experience block rendered as a two-column table. The Mercor intake parser extracts 'Senior Researcher' as a skill instead of a job title.

Resume parse pipeline must distinguish job-title sections from skills/keywords blocks, preserve role-employer-dates triples, and surface a confidence score per extracted field. Low-confidence fields should be flagged for candidate confirmation in the intake UI rather than silently dropped into a ma…

A customer lab opens a project requiring 'mathematical biology + RLHF rubric authoring experience.' The semantic matcher returns candidates whose only signal is the word 'biology' in their resume.

Matching must require AND-style satisfaction of the conjunction of required skills, not OR-style keyword overlap. Surface per-candidate evidence spans for each required skill so operators can audit the match. [REQUIRES-VERIFICATION] on Mercor's published precision/recall metrics for the matcher.

A candidate re-applies under a slightly different email (gmail vs gmail with dot) and a typo'd legal name. The matcher treats them as two distinct candidates and both end up in the same lab's pool with conflicting ratings.

Dedupe at intake using a combination of normalized email (Gmail dot/plus rules), phone, government-ID hash (post-KYC), and resume fingerprint. On suspected match, merge profiles and reconcile ratings — do not deliver two copies of the same person to a customer lab.