All case studies
RecruitmentAI MatchingBlueprint

The right candidate was
already in your database.

A talent matching engine that reads every resume an agency owns, turns each new job description into a real requirement, and returns a ranked shortlist with the reasoning attached. People still make every call.

Industry
Recruitment & Staffing
System
Talent matching & screening engine
Built on
Resume parsing · Hybrid matching · WhatsApp AI
Status
Blueprint · ~4 week build

The problem

Ten thousand resumes,
and none of them findable.

A recruitment agency’s most valuable asset is the database it already owns. Yet when a new role lands, the fastest path is almost always a fresh job post, because nobody can search what’s in there. Job titles are inconsistent, half the older files are scanned PDFs, and the same person exists four times under three email addresses.

So the agency pays twice: once to re-source candidates it already had, and again in the placements it loses to whoever submits a good CV first. Speed to shortlist is how agencies win, and manual screening is exactly where it gets lost.

The system

Read everything. Rank with reasons.
Let people decide.

Every resume lands in one place

Email, job boards, WhatsApp, referrals. Each one is parsed into a structured profile: skills, titles, years, employers, city, notice period, expected salary. Scanned files go through OCR, so the 2019 PDFs stop being dead weight.

One person, one profile

Four applications, two email addresses, three years apart, collapse into a single candidate with a history. No more calling someone about a role they were rejected for last spring.

The job description becomes a requirement

A client sends a messy Word doc; the engine pulls out must-haves, nice-to-haves, budget ceiling, location and notice window. The recruiter confirms it in thirty seconds instead of retyping it.

Matching runs both directions

A new brief ranks the entire database; a new resume checks itself against every open role across every client. Meaning-based search finds the people whose titles don’t match but whose work does, while hard filters hold the line on budget, city and notice period.

Dormant candidates come back

People who applied last year and now fit get surfaced automatically, with a WhatsApp message that mentions the specific role. A database that used to age into nothing starts producing placements.

AI pre-qualifies, a human decides

Notice period, salary, location and interview availability are confirmed on WhatsApp before a recruiter spends a call. The shortlist then goes out branded and contact-anonymised. Nothing is ever auto-rejected.

Watch it work

One brief in,
a defensible shortlist out.

New brief from clientSenior Backend Engineer · Pune · ₹18–24L · 30-day notice
Scanning the database
1
R. NairBackend · 8 yrs · Pune
5/5 must-haves 30-day notice In Pune
92
2
A. DesaiBackend · 6 yrs · Bengaluru
4/5 must-haves 60-day notice Open to relocate
81
3
M. KulkarniBackend · 9 yrs · Pune
9 yrs experience No cloud cert — ask on call
74

Quick check before we submit you: what’s your notice period?

30 days

Shortlist ready · awaiting recruiter approval

Illustrative run. The names, scores and timings show how the system behaves, not a specific agency’s data.

What it changes

The database starts earning.

100%

Of the existing database searched on every new brief, including the resumes that arrived three years ago.

Minutes

From job description to a ranked, explainable shortlist, instead of an afternoon of manual screening.

0

Candidates auto-rejected. The engine ranks and explains; a recruiter approves every shortlist that leaves the building.

Guardrails

AI that ranks.
People who decide.

Identity-blind scoring

Name, gender, age and photo are stripped before anyone is scored. The engine ranks on evidence: skills, years, and what a person has actually shipped.

Nothing is auto-rejected

The engine sorts and explains. It never closes a door. Every shortlist that reaches a client has a recruiter’s name attached to it.

Every score is auditable

Each ranking keeps the reasons that produced it, so a decision made in March can still be explained in September, to a client or to a regulator.

Consent and retention built in

Candidates know they’re in the database, and profiles age out on a schedule you set instead of living in a folder forever.

How we’d build it

Four weeks, in the order that matters.

  1. 01

    Map the intake

    Where resumes actually arrive: inboxes, job boards, WhatsApp, referral spreadsheets. And what “a good match” means for each client you place into, in their words.

  2. 02

    Build the parse and dedup layer

    Every format, including scanned PDFs, into one structured profile per human, with the application history kept intact.

  3. 03

    Turn matching into scorecards

    Meaning-based ranking combined with hard filters, and the reasoning surfaced as plain chips a recruiter can defend on a client call.

  4. 04

    Put the pre-qualifier in front

    WhatsApp or voice confirms notice period, salary, location and availability, so nobody burns a screening call on a candidate who was never available.

  5. 05

    Ship the shortlist pack and pipeline

    Branded, anonymised summaries out to clients, and submitted to interviewed to offered to joined tracked in the CRM with automatic nudges on both sides.

Sitting on a database you can’t search?

One free call. Bring a live job description and we’ll show you what a ranked shortlist would look like.

Next case studyMedical Clinic AI Receptionist