Extract Resume Information with AI
Candidate name, current role, years, skills, education — résumés into consistent screening records, with inference kept on a short leash.
Overview
Résumés bury the same facts in a thousand layouts. Screening at scale needs the same fields out of every CV: candidate_name, current_role, years_experience, skills as a deduplicated list, education, location. The catch is inference — "is the top role the current one?" — so this setup runs conservative: the model may take one short step from the text (the role dated "2024–present" is current) but never a chain of assumptions. Missing values come back empty rather than null, keeping the record human-readable in a screening sheet.
Workflow
-
Generate and feed one résumé per call
One CV per extraction keeps the example shape honest and the fields unambiguous.
-
Check the reading guidance
"Most-recent-first is common but not guaranteed — check the dates." That line saves the current_role field.
-
Keep conservative inference
Years of experience may be summed from dates (one short step); seniority may not be invented from job titles.
Why This Works
- Fixed fields make a thousand layouts comparable in one sheet
- The skills list rule (one entry per item, deduplicated) ends comma-blob prose
- Conservative policy permits the date-math inference screening needs while blocking speculation
Best for
- High-volume screening where layout chaos kills comparisons
- Recruiters who need records, not summaries
- CVs where the current role isn't obvious from position alone
Not for
- Scoring or ranking candidates — that's judgment, not extraction
- Deciding fit against a job description — pair with a Role Prompt for evaluation
Use cases
- Building a screening sheet where every candidate has the same columns
- Getting skills as a clean, deduplicated list instead of prose
- Computing years of experience as a digit, not "over a decade"