AI Matcher turns CVs and job ads into structured data, ranks people against roles in any language, and tells your recruiters why each match scored the way it did. One REST API. The API and its database are hosted in the EU, or the whole stack runs on your own servers.
Positions with dates and employers, education, skills, languages, certifications and contact details, as clean JSON. Values that are not in the document are dropped, never invented.
Rank the candidates in your database for a job, or the open jobs for a candidate. Filter on tags and document ids, and search several databases at once.
Recruiters see which titles, skills and languages matched, and can weight categories their own way. Scores are not learned from your past hiring decisions.
POST /cvs/parseReads multi-column layouts, scanned pages and old Word files. Strips text that tries to instruct the AI and flags the CV for review.
POST /jobs/parsePulls the title, required and preferred skills, languages, education and years of experience out of any job ad.
POST /search
POST /indexes/…/matchesSemantic ranking across your candidate databases, with filters on tags, document ids and custom values.
GET /indexes/…/scoringA language model reads each CV against each job ahead of time and stores a fit score with category scores, ready the moment a recruiter asks.
POST /cvs/previewFill your create-candidate form with name, email and phone in seconds; the full profile is ready by the time the recruiter clicks save.
Index overviews, rescoring, and a GDPR erase that removes a candidate from every index with a receipt you can keep.
Two ways of scoring work together, so a recruiter never waits for the AI and never gets a half-finished list.
Each CV and job is parsed once and turned into vectors for its full text and for every skill and title.
A match answers in under a second from the stored vectors, across languages, with no model call on the way.
A language model scores every candidate for each active job, closest first, and keeps the scores.
Once every candidate of a job is scored, the list answers with AI fit. Lists are never a mix of the two.
The API speaks the request and response shapes most recruitment software already uses for parsing and matching. Point your existing integration at a new base URL, keep your index and document ids, and move your data across with our migration tools.
POST /api/v1/indexes/jobs/documents/job-42/matches
X-Api-Key: aim_••••••••
{
"indexes": ["candidates"],
"limit": 20
}
200 OK
{
"ranked_by": "ai_fit",
"matches": [
{ "id": "cv-1093", "score": 92,
"ai_fit": { "score": 92,
"categories": { "job_titles": 95, "skills": 90 } } }
]
}
The API and its database run in the EU. Larger customers can run the whole stack, models included, on their own servers.
Parsing and scoring use open models on dedicated GPUs. CVs are never sent to a third-party AI provider.
Every environment's indexes are kept apart. One customer can never search, score or delete another's data.
Remove a candidate from every index, with their scores and queued work, and keep a PDF receipt.
Hidden instructions in a CV are removed before any model reads it, and the CV is flagged.
The AI reads a CV without name or contact details, and scores against the job's requirements only.
Up to 1,000 CVs and 20 jobs of your own, ranked side by side with your current results.
Start a pilot10,000 CV and job parses a month, matching and AI fit scoring included, one production environment. Parsing pauses at the monthly limit; upgrade any time.
Talk to usSeveral environments, on-premise deployment, migration of your existing indexes, and volume pricing.
Contact salesPDF, DOCX, DOC, RTF, ODT and plain text, plus PNG, JPEG, TIFF, BMP and WEBP images through OCR. Scanned PDFs are read the same way. A file can be up to 10 MB and a PDF up to 30 pages.
CVs and job ads are read in their own language, and every skill and title is also mapped to an English term, so a Dutch CV can match a French job. We test on English, French and Dutch; bring CVs in your languages to the pilot and check them yourself.
Each CV or job ad you send for parsing counts once. A quick preview followed by the full parse of the same file counts once. Searching, matching, AI fit scoring, tag updates and deletes are included and never counted.
The API uses the request and response shapes most recruitment software already speaks, so your integration keeps working with a new base URL. Our migration tool copies your existing CVs and jobs under the same index and document ids, can be stopped and resumed, and catches up on changes before you switch. Switching back is the same one-line change.
The API and its database run in an EU data centre. The AI models run on our own GPUs: CVs are never sent to a third-party AI provider. Each customer environment is stored apart, and no other customer can search, score or delete your data.
Yes. One call removes the CV from every index of your environment, together with its scores and any queued work, and gives you a receipt with the date and the indexes it was removed from.
It reads the CV without the name or contact details and compares it with the job's requirements: titles, skills, languages, education, certifications and level. You get an overall score and a score per category, and you can weight the categories your own way. It does not learn from your past hiring decisions.
Accuracy depends on your roles and your candidates, so we measure it on yours. In the pilot, recruiters rate the top candidates of AI Matcher and of your current system side by side without knowing which is which.
Yes, on the Enterprise plan. The whole stack, models included, runs on your hardware; you need a server with a recent NVIDIA GPU with 32 GB of memory or more.
Every plan includes email support from the engineers who build AI Matcher, and full API documentation with example requests and responses.
Send us a sample of CVs and jobs. We run them through AI Matcher and walk you through the results.