Smart Match HR

Shortlist senior roles on evidence, not keyword luck.

Smart Match HR reads a job description and a folder of CVs, ranks every candidate 0 to 100 across skills, experience, education, and role relevance, and links each score back to the exact line of the CV it came from. Built for talent teams drowning in applicants for senior and specialist roles.

Every score traces to quoted evidence in the source CV, so any recruiter can audit a ranking in seconds.

Enterprise Data AI — Semantic Layer

The Problem

Senior roles get buried, and the best candidates leak through the cracks.

Recruiters spend the day re-reading the same PDFs, scanning for keywords a parser already missed, then ranking by gut. The signal is in the CV. The time to find it is not.

250+ Applicants per senior role Median, EU tech market 2026

23h Recruiter time per shortlist For an 8-person shortlist

7s Average time spent per CV Long enough to miss the signal

“I’m not screening, I’m pattern-matching strings. By Friday I can’t tell two candidates apart, and the best ones leak through the cracks.”
Talent Lead, Mid-Stage SaaS

The Solution

The semantic read a recruiter would do, done in minutes and shown its work.

Smart Match HR does the matching a strong recruiter does in their head, then explains it. It maps what a job actually requires to the evidence in each CV, so “Spring Boot” matches “Java Enterprise” instead of failing a literal keyword check. The output is a ranked leaderboard you can sort, filter, and defend.

Who it is for

Talent teams hiring senior or specialist roles where one posting pulls hundreds of applicants and the difference between candidates is in the detail, not the headline.

Who it is not for

Low-volume hiring where a recruiter already reads every CV end to end. If you post a handful of roles a quarter, you probably do not need this yet.

Capabilities

Six things the tool does the moment you drop in a folder.

Drop a folder, get structure

Parse a mixed stack of PDF, DOCX, and plain-text CVs against one job description into a clean, structured read. No reformatting, no manual data entry.

PDF · DOCX · TXT

Matches meaning, not strings

Job requirements map to candidate evidence by meaning, so “Spring Boot” matches “Java Enterprise” and strong candidates stop failing on vocabulary.

JD ↔ CV embeddings

A score you can change

A weighted 0 to 100 score across skills, experience, education, and role relevance. Every weight is editable, so the ranking reflects how your team actually hires.

explainable · auditable

Evidence, not vibes

Every claim links back to the exact line of the CV it came from. Audit any ranking in seconds instead of taking the score on faith.

source-linked

A workspace, not a report

Sort, filter, pin, and reject from a ranked leaderboard built to be the daily screening surface, not a one-off PDF.

React workspace

Name and photo redacted while scoring

Candidates are ranked on merit during scoring, with name and photo removed, so identity does not drive the ranking.

fair · auditable

The Pipeline

Six agents, one pass from folder to leaderboard.

The work runs as one automated pipeline. You drop in the JD and the CVs; the leaderboard comes back ranked with evidence attached.

01 Ingest

Input Agent

Takes in the JD and CV files, normalizes the formats, and routes them downstream.

02 Parse

Parsing Agent

Extracts skills, roles, education, and dates into a typed schema.

03 Embed

Skills Intelligence

Reads synonyms and equivalents, mapping “Spring Boot” to “Java Enterprise”.

04 Match

Matching Agent

Aligns each JD requirement with the evidence in each candidate’s CV.

05 Score

Scoring Agent

Computes the weighted 0 to 100 score across the four axes, with editable weights.

06 Rank

Output Agent

Returns a ranked leaderboard with quoted evidence behind every candidate.

Fairness and Auditability

Every ranking comes with the receipt.

Screening tools earn trust by showing their work. Smart Match HR removes name and photo before scoring so identity does not influence the rank, and links every score back to the line of the CV behind it.

Name and photo redacted during scoring, so identity does not drive the ranking.
Every score links to the exact CV line behind it, ready to audit or overrule.
When a hiring manager or candidate asks why someone ranked where they did, the answer is on screen, not in a recruiter’s memory.

Connections to source systems run over MCP (Model Context Protocol), so data access is governed rather than ad hoc.

Where It Stands

A working pipeline today, integrated into your stack next.

Live today · V0.1

What runs end to end now

End-to-end pipeline: JD upload to ranked leaderboard
Semantic JD-to-CV matching with strength scores
Four-axis weighted scoring with editable weights
Per-skill quoted evidence from source CVs
Bias guardrails: name and photo redacted
React workspace: sort, filter, pin, reject

Planned · next 30 days

From spike to internal pilot

Job descriptions pulled from Microsoft Dynamics
CV pool pulled from BambooHR
Internal pilot with a live recruiting team
Scoring refinement from recruiter feedback
Expanded CV-pool ingestion pipeline

The Pilot

See it rank a req you already closed.

The fastest way to judge a screening tool is to point it at a role you have already hired for and see whether it would have surfaced the person you actually picked. A scoping call sets up a pilot on your own job descriptions and a sample of your own CV pool, with your scoring weights, so you are evaluating it on your hiring, not a demo dataset.

The pilot covers your roles, your CVs, your weights, and a side-by-side against how the shortlist was built manually.

FAQ

Straight answers before the call.

Is this going to automate our existing bias?

Name and photo are removed before scoring, and every score links to the specific CV line behind it, so a person can see exactly what drove a ranking and overrule it. A formal bias-audit process is part of what a pilot would establish.

How is this different from the matching in our ATS?

Most ATS matching scores on keyword overlap. Smart Match HR scores on meaning, so equivalent skills phrased differently still match, and it shows the evidence behind each score rather than returning a number with no explanation.

Who maintains it after the pilot?

That is part of what the scoping call defines. The current build is a prototype; a pilot establishes the ownership and support model before any rollout.

What happens to candidate data?

Connections to source systems run over MCP, so access is governed. Formal data-residency and retention terms are set during pilot scoping.

How fast would we see something usable?

The pipeline already runs end to end today. A pilot points it at your own roles and CVs; the planned Dynamics and BambooHR connections are targeted for the following 30 days.

What does it cost?

No pricing is set yet. The first step is a scoping call to size a pilot against your hiring volume.

Get Started

Stop ranking senior candidates by Friday-afternoon gut.

Bring one role you have already closed. We will run Smart Match HR against it and show you the ranking, with the evidence behind every score, in a single call.