Resume parsing and deduplication across 60,000 profiles

Project Description

Our database has 60,000 resumes accumulated over six years, in PDF, Word and a few as photographs. The same candidate appears up to five times under different email addresses. We want structured extraction — skills, companies, dates, current location — and a deduplication pass that merges without destroying, because a wrong merge loses a candidate history. Recall matters more than precision on the parsing; a recruiter can correct a field but cannot find a candidate who was never indexed.

Skills Required