The problem
A UK-based recruitment firm placing 200+ candidates per month had recruiters spending 60% of their time on first-pass CV screening. For each role, a recruiter would manually read 80–150 CVs, score them against job criteria on a spreadsheet, and create a shortlist. It was slow, inconsistent between recruiters, and deeply unsatisfying work.
The pipeline we built
1
Multi-source CV ingestion
n8n workflow monitors email inboxes, LinkedIn message inbox, job board APIs (Indeed, Reed, Totaljobs), and ATS uploads. CVs in any format — PDF, DOCX, HTML — parsed to structured text using a Python extraction layer with fallback OCR for scanned documents.
2
Structured candidate profile extraction
An LLM extracts structured data from raw CV text: years of experience by skill, education level and institution, employment history with tenure calculations, certifications, and location. All stored as structured records in PostgreSQL — fully queryable and filterable.
3
Per-role scoring engine
Each job specification is parsed into weighted scoring criteria. The LLM evaluates each candidate against the criteria and returns a 0–100 score, a confidence level, and written reasoning for the top 5 scoring factors. Recruiters can adjust criteria weights without code changes.
4
Bias mitigation layer
Names, gender indicators, graduation years (age proxy), and non-professional social links are stripped before scoring. The LLM sees only relevant professional information. Fully auditable — every scoring decision is logged with the sanitised input.
5
ATS shortlist push
Candidates scoring above the configurable threshold are pushed directly to the ATS (Workable, Greenhouse, or Teamtailor) with the AI score and reasoning attached. Recruiters review shortlists, not raw CVs. Average time-to-shortlist dropped from 3 days to 4 hours.
We used to tell clients it takes 3–4 days to get them a shortlist. Now we send it the same day. Our recruiters are spending their time on calls and placements — not reading CVs. It's the highest-impact tool we've ever deployed.
— Managing Director, Recruitment Firm, UK
