How is this different from keyword matching in an ATS?

An ATS filter looks for exact words and discards anyone who phrased things differently. Talenture scores the underlying skills and their relationships, so a candidate with transferable experience still surfaces. You rank by fit, not by who happened to use the right keyword.

What signals does the match score include?

Eight: direct skills, transferable related skills, semantic similarity, experience, seniority and trajectory, location and work type, company values, and salary fit. Each is weighted for the role and combined into one score you can break down per candidate.

How does semantic matching work?

Talenture turns each profile and role into a vector embedding and compares them with cosine similarity, so it recognises when two profiles describe the same strength in different words. A similarity gate blocks weak pairings, for example two unrelated skills that happen to share a broad domain.

Does it account for location and salary?

Yes. Commute distance is scored for onsite roles and eased for hybrid and remote, with a fair floor for candidates open to relocation. Salary expectations are checked against the role's range, and a candidate just outside a rule can still be shown for hiring-manager flexibility.

Can I see why a candidate received their score?

Yes. Every match breaks down into requirements coverage, matched and missing skills, experience, location, values, and salary signals, with the specific evidence behind each one. The reasoning is in plain language, so you can justify any ranking to a hiring manager.

What data is the matching built on?

A skills graph of 4,600+ verified skills and 21,700+ AI-validated relationships across 240+ tech roles, built on ESCO v1 (EU) and O*NET (US Department of Labor). It understands progression, transferability, and semantic similarity between skills.

Does it work for non-technical roles?

The graph is strongest for Data, AI, and Engineering roles, which is where Talenture focuses. It still scores adjacent and operational roles, and coverage keeps expanding as more roles are mapped.

Will the matching introduce bias?

Scoring is based on skills, experience, and values, not demographic signals. Because every score is explainable, you can audit exactly why any candidate ranked where they did, which is far harder with a black-box tool.

Stage 2 of the hiring mission · Find and understand candidates

Two hundred applicants share the job's keywords. Three can do the job.

Talenture scores every applicant across 8 signals and shows you exactly why, so when the hiring manager asks "why this candidate?", the answer is on one page.

Works with your ATS: candidates flow in, scorecards flow back.

Talenture ranks the applicants your role attracted, then hands the strongest to the preparation step.

See a sample scorecard Book 15 minutes with the founder

Every: applicant scored, not just the first twenty
8: signals behind every score
One page: of evidence per candidate

What is AI candidate matching?

AI candidate matching scores how well each applicant fits a role using direct and transferable skills, semantic similarity, experience, seniority, location, company values, and salary, instead of keyword overlap. Talenture weighs these eight signals into one score, ranks your pipeline automatically, and shows the evidence behind every result.

The match layer of the candidate scorecard: a defensible fit score, plus the interview playbook of gaps to probe.

See the full scorecard

app.talenture.ai/roles/platform-engineer/applications

Live demo

This is the real product running on sample data, not a screenshot.

Every applicant ranked, with the eight signals behind the score on the candidate.

Eight signals, weighted for the role

Most tools match strings. Talenture scores each applicant across eight signals at once and weights them for the role, so a strong candidate is never lost to a missing buzzword.

Direct skills: the ones your role actually names.
Transferable skills: a React developer can be a strong Vue hire.
Semantic fit: the same strength, described in different words.
Experience: how much of it is the work you are hiring for.
Seniority and trajectory: career direction, not just years served.
Location and work type: commute scored, relocation kept in the running.
Company values: culture as a scored signal, not a gut call.
Salary: expectations checked against your range, up front.

Open one candidate

The scorecard a recruiter reads: the verdict, the score breakdown, and a rail to the evidence, the next steps and the interview plan. Click through it.

Pre-screen rules and salary, made visible

Hard requirements on seniority, location, and salary are applied up front, and the result is never hidden. When a candidate sits just outside a rule but is worth a look, they are shown for hiring-manager flexibility with the reason attached.

Must-have rules on seniority, location, and salary
Salary expectations checked against the role's range
Borderline candidates shown for flexibility, with the reason

Then check the working

Every signal scored separately, with the evidence that produced it, so the number survives a hiring manager asking why.

Every ranking is defensible

When a hiring manager asks why a candidate ranked where they did, you have the answer per candidate: which requirements they cover, which skills matched and which are missing, the reasons behind the score, and the gaps to probe in an interview.

Requirements coverage, matched and missing skills
Plain-language reasons behind every score
Gaps to probe, ready for the interview

How AI candidate matching works

Four analytical layers run in parallel for every applicant, each producing a partial score that feeds the overall match.

1
Skills graph: credit for transferable and related skills
A graph of 4,600+ verified skills and 21,700+ AI-validated relationships identifies substitutes, overlaps, and prerequisites so candidates with transferable experience surface instead of being buried by literal filters.
2
Semantic matching: meaning over exact wording
Vector embeddings compare the meaning of each candidate profile against the role using cosine similarity, so the same strength described in different words still registers. A similarity gate blocks weak pairings inside a shared domain.
3
Seniority and trajectory: career direction, not just years
The engine reads career progression, not only tenure. A logical next-step is rewarded, overqualification is flagged rather than silently ranked first, and a fast riser is not penalised for a short title history.
4
Location scoring: real commute distance, not a radius filter
Commute distance is calculated from real coordinates for onsite roles, eased for hybrid and remote, and candidates open to relocation receive a fair floor score instead of a hard exclusion.

Frequently asked questions