---
title: "AI candidate and client-need matching platform · Client case · Quentin Andrieux"
description: "Engineering consultancy: candidate, client and recruitment spaces, CVs read and matched by AI against the firm's own skills framework. Anonymised client case."
url: https://quentinandrieux.com/en/client-cases/candidate-matching-platform/
lang: en
author: Quentin Andrieux
updated: 2026-11-11
---

> Version texte de https://quentinandrieux.com/en/client-cases/candidate-matching-platform/ (the HTML page is the reference). Author : Quentin Andrieux, https://quentinandrieux.com/

[Home](https://quentinandrieux.com/en/) [Client cases](https://quentinandrieux.com/en/client-cases/) Matching platform

Engineering consultancy · recruitment and placement

# A matching platform built on the firm's own framework.

A firm that places engineers with industrial clients wants to handle more inbound requests and a larger business volume. AI reads the CVs, maps them to its skills framework and matches profiles with client needs.

**15** field interviews before the first line of code: business managers, consultants, clients

**≈ 1,100** skills in the framework the AI uses to classify each CV

**1 to 4 cents** the cost of analysing one CV, measured in trials

The flow

1.  **CV uploaded** The profile fills itself in, then availability and mobility.
2.  **Classification** Experience and skills mapped to the firm's framework.
3.  **Technical exercise** Trade-specific questions, graded by the AI.
4.  **Matching** Profiles matched to needs, by the firm's rules.
5.  **Recruiter** Assesses, moves the stage; follow-ups go out on their own.

From the CV to a ranked profile for a client need.

## The context

The firm places engineering consultants with industrial clients. It wants to handle more inbound requests and a larger business volume. Fifteen field interviews set the diagnosis: business managers juggling CRM, spreadsheets and billing tools, a database search limited to one keyword, skills files reformatted by hand, applications left waiting.

## What was built

- **Candidate space**: one link per job offer; the uploaded CV fills in the profile, then availability, mobility and a technical exercise specific to the trade. The candidate leaves with a skills passport and can follow their stages.
- **Client space**: the need expressed through a guided flow, which becomes a structured requirement sheet.
- **Recruitment space**: a six-stage application tracker, rules for follow-ups and stage changes, a ranking of profiles per offer, the recruiter's assessment, the history of every change, duplicate detection.
- **The website**, which the team edits on its own.

## What the AI does

- It reads a CV and maps experience and skills to the firm's framework (around twenty trades).
- It asks and grades technical questions per trade, with a constrained vocabulary approved point by point by the firm.
- It matches profiles and needs according to weighting rules set by the firm: major or minor gap, CV freshness, recruiter's assessment.
- It sources external profiles, after benchmarking nine tools, for about one cent per profile.

Ranking for a need · methods engineer *Fictional example*

#### Profile 1 · 92%

Trade match

Skills 14/16

CV freshness 8 months

Minor gap: mobility

#### Profile 2 · 74%

Trade match

Skills 9/16

CV freshness 1 month

Major gap: certification

Fictional data. Weighting rules are set by the firm.

## How

One month of field work and prioritisation, then a build in blocks delivered every week and reviewed with the project lead: candidate intake, client space, website, recruitment tracker, matching. Code, hosting and database were transferred to the firm's own accounts.

## Where the project stands

The recruitment tracker has been live since early November and the first real candidates go through the platform. Matching relevance is being validated on real placements, with an evaluation grid proposed by the firm.

## Further reading

- [Sales cockpit connected to the ERP](https://quentinandrieux.com/en/client-cases/sales-cockpit-erp/)