Generative AI · Human Resources
AI-powered ATS: read every CV and rank candidates with judgment, not keywords
An attractive opening receives hundreds of CVs, and the classic keyword filter discards valuable people over a wording detail. We built a recruiting system where AI reads every CV in full, understands the experience even when described in different words, and delivers a ranking with the reasons in plain sight.
Problem
Hundreds of CVs per opening: keyword filters discard good candidates who described the same experience in different words.
Result
Every CV read and structured by AI, and an explained ranking: why each candidate stands where they stand.
The context
The traditional recruiting funnel has a known defect: nobody can read five hundred CVs, so they filter by keywords — and the candidate who wrote "built interfaces with React" passes, while the one who wrote "developed the platform’s frontend" is lost, even if they are better.
The result is two invisible costs: good candidates discarded over wording, and recruiter hours spent reading CVs a better filter would have sorted. In markets where talent is scarce, the first cost is the expensive one.
How it works
The system receives CVs in any format — PDF, Word, scans — and turns them into structured profiles: real experience, technologies, achievements, trajectory. It does not look for words: it understands descriptions. Experience with a technology counts even if the CV never uses its exact name.
Against the role description, the engine combines semantic understanding with the explicit requirements and produces a ranking where every position is argued: what this candidate has, what they lack, where their CV says it. The recruiter does not receive a score — they receive a case.
Engineering decisions
Decision 01
Explained ranking, not a black box
Every score comes with its reasons, citing the CV. In a decision that affects people, a number without an argument is not acceptable — not for the candidate, and not for the recruiter who has to defend their shortlist.
Decision 02
Semantics over keywords
Experience described in different words counts the same. Matching combines embeddings — closeness of meaning — with the role’s hard requirements, instead of counting literal matches.
Decision 03
AI sorts; the human decides
The system never discards anyone on its own: it prioritizes and argues. The decision to advance a candidate or not belongs to the recruiter — with better information, in a fraction of the time.
What changes for the business
The change is not just speed: the shortlist becomes defensible. Every candidate who advances — and every one who does not — has documented reasons.
- From hundreds of CVs to an argued shortlist in minutes.
- Fewer false discards: wording stops eliminating good candidates.
- Auditable process: every decision has its reasons in writing.
- Recruiters spend their time interviewing, not filtering.
Where else it applies
Corporate HR
High-volume openings with hundreds of applicants.
Headhunters & staffing
More simultaneous searches with the same team.
Technology
Technical profiles where keywords mislead the most.
Universities & programs
Scholarship and admissions selection with traceable criteria.
FAQ
Frequently asked questions about this case
How do you handle bias risk in selection?
Three safeguards: the role’s criteria are explicit and configurable, every score comes with an auditable explanation citing the CV, and the system never discards on its own — the final decision is always human.
What CV formats does it accept?
PDF, Word, plain text and scanned documents. The pipeline includes OCR, and the model understands diverse structures: candidates do not have to adapt their CV to the system.
Does it replace our current ATS or complement it?
Both paths work: it can operate as a complete system or integrate into your current flow, contributing the reading and ranking layer. It depends on how much you want to change at once.
Does it work with CVs in Spanish and English?
Yes, and with mixed batches — common in technical searches in Latin America where CVs arrive in both languages.
Have a similar challenge?
Tell us your problem. We'll tell you how to solve it — and how long it takes.