For recruiters
Screen candidates against one vacancy
Paste the text of a vacancy. Victra scores it against people who uploaded a CV here and agreed to be found, and returns them ranked with the reasoning attached — what fits, what is missing, and how far the experience is from what the posting asks for.
How a run works
You paste a vacancy — between 30 and 20 000 characters — and the assistant parses out the role, level, experience, location and skills, all of which you can correct before the first run. It then runs on a schedule and re-runs when you change the text. Up to ten screening assistants per account, each with its own vacancy and its own threshold.
Who is in the pool
People who registered here, uploaded a CV, and agreed to appear in candidate matches. Nobody else. We do not scrape named strangers off the public web and rank them as candidates: it produces lists of people who will never reply, and it processes personal data without any basis for doing so. That sourcing exists in the code and is switched off.
How the score is computed
Two phases, because reading every CV with a model would cost more than it is worth. First a keyword overlap between the vacancy and each CV ranks up to 300 candidates, most recently updated first. The top 30 are then read by the AI, which scores each one and writes out the strengths and the gaps. The scoring rules are the same ones used to score vacancies for candidates, so a percentage means the same thing on both sides of the platform. You set the threshold — anywhere from 10 to 100, sixty by default — and anything below it is kept but not shown until you ask for it.
What you see, and what stays hidden
Role, field, seniority and years of experience are always shown, because that is what a ranked list is for. A name, a photo, an email, a phone, a LinkedIn or GitHub link appear only when that candidate has opted in to being visible to employers, or has accepted your request to be contacted. The CV opens under the same rule, and when it does not open the reason is named — consent not given, no file, a public profile with no document — because "no CV" and "we are withholding one" call for different actions. A candidate who applies to your own vacancy has handed over their contacts by applying; there is nothing to request there.
Your own vacancies, and their applications
A company that claims its page can publish vacancies into the public catalogue and receive applications in one inbox. Each application moves through stages you set — new, reviewed, shortlisted, interviewing, offered, hired or rejected — and the candidate sees the stage they are at. A rejection can carry a short note, and they see that too.
What it does not do
It does not schedule interviews, keep interview notes, or write to candidates on your behalf. It does not reveal a contact the candidate has not agreed to share. And the screening tools are not switched on by registering a company: the recruiter role is requested and approved by hand, because it opens other people’s CVs.
Questions
Where do the candidates come from?
From people who signed up on Victra, uploaded a CV and agreed to be found by employers. There is no scraped database behind this, and no profiles imported from elsewhere.
Can I see a candidate’s name and contacts?
Only when they have made themselves visible to employers, or when they accept your contact request. Until then you see the role, the level, the years of experience and the AI reasoning — enough to decide whether to ask.
Is the match a keyword search?
No. Keywords only pre-rank the pool so the AI reads the most plausible thirty CVs rather than all of them. The score itself comes from the model comparing the CV with the vacancy, and it explains what fits and what does not.
How do we get access?
Claim your company page, then request the recruiter role from your profile. Approval is manual — a person reads the request — because the role opens documents that belong to other people.