Candidate matching for recruitment teams
Match candidates to jobs using structured recruitment data
Turn a CV and job description into structured records, then inspect the components behind the match. Recruiters keep the final decision.
How a match is built
Job description
Parse the role into structured requirements, including skills and experience fields.
Candidate CV
Parse the candidate into a profile with skills, experience, education and other available fields.
Reviewable match
Use the total score, component scores and evidence as a starting point for human review.
See more than one opaque score
The documented response contains a total score and separate component fields. The scoring service applies recruiter-configured weights to criteria with usable data; the response can also include evidence and the algorithm version.
- Profile (occupation)
- Compare the role and candidate occupation titles and taxonomy evidence.
- Skills
- Compare extracted candidate skills with job skills.
- Experience
- Review experience alongside the role requirements.
- Education
- Compare the candidate's education with the job requirements when both are available.
- Languages
- Compare specified language requirements with candidate language data.
- Distance
- Review the location-based distance component when location data is available.
- Semantic match
- Compare role and profile text using stored embeddings when available, with a lexical fallback.
- Career trajectory
- Review recent role history in relation to the target role.
A score helps prioritize review; it does not decide whether to hire, reject or contact a person. Confirm the role criteria and examine the underlying CV before acting.
Matching in your ATS or application
The API documents stored-resource matching and a file-to-file matching endpoint, along with a response containing score components and evidence. Teams can keep the review inside their existing workflow.
Explore a synthetic example
Download the CV inputs and captured API outputs. All people, employers and vacancies are fictional. The sample documents are in English; results depend on the input and parser and require human review.