Four agents read one candidate at once
The résumé, the LinkedIn record, the GitHub history and a live voice interview — assessed in parallel, against a rubric built from your own job description. What lands on your desk is a one-page brief with the evidence behind every score, and the decision still yours to make.
4
agents on one candidate, running in parallel
1
rubric, applied to every applicant in the same depth
0
hiring decisions made by the system
Screening is the slowest step, and the least consistent
Not because anyone is careless. Because a human first pass is a finite resource being spread across an infinite pile.
Nobody reaches the bottom of the pile
Good candidates are missed for the most ordinary reason there is: the pile was long, the day was short, and the first fifty got the attention.
The same CV scores twice
Two reviewers read the same application and reach two conclusions, and the reasoning behind a rejection is rarely written down anywhere it can be checked later.
Keyword filters reject people who can do the job
The blunt instrument used to make the pile smaller is the one least able to tell whether somebody can actually do the work.
Four readings of the same person, at the same time
Parallel rather than sequential. That single structural choice is the whole speed story, and it is why a candidate hears back the same day.
The résumé
Read against a rubric generated from your own job description, not against a keyword list. The point of the rubric is that it exists before the first application arrives and does not drift as the reader gets tired.
The professional profile
The LinkedIn record: the trajectory, the tenure, the recent moves, the things a CV formats away. Public data only, and only with the candidate's consent at collection.
The code
For engineering roles, the GitHub history — what they actually build, how they build it, and how consistently. The single most under-read signal in technical hiring, because reading it properly takes an engineer an hour.
The conversation
A live AI voice interview that asks follow-up questions rather than reading from a card, in the candidate's own language and their own timezone, so the first real conversation is not scheduled three weeks out.
One page per candidate, and every score shows its work
A score with no evidence is an opinion with a number on it. Each one here links back to the line, the commit or the answer it came from.
The scores
Against your rubric, criterion by criterion, in the same depth for every applicant who applied.
The evidence
Each score links to what produced it: the specific line in the CV, the specific repository, the specific thing the candidate said.
The recommendation
A stated conclusion your hiring manager can read in a minute and disagree with in two — which is the only kind worth writing.
The brief is also the record. Six months later, the reason a candidate was or was not advanced is written down, with what it was based on — which almost never survives a manual first pass.
Your applicant tracking system stays the source of truth
VectorHire plugs into the ATS you already run and writes its results back into it. A screening tool that demands a migration first does not survive contact with a recruiting team.
Candidates come in
From the job boards and the careers page you already use, with consent captured at collection.
Screening happens beside
The four agents work, the brief is produced, and interviews are scheduled against your team's real availability across timezones — nothing about your existing pipeline changes shape.
Results go back
Candidates, scores, interview outcomes and status write back where your recruiters already look.
We will not tell you this system is unbiased
Nobody can demonstrate that about an AI system, and hiring is the one category where the claim is both unprovable and regulated.
What we will tell you is what the system does, which is checkable: it applies the same defined criteria to every applicant, in the same depth, and shows the evidence for every conclusion so a person can see it and overrule it. Automated employment decision tools are regulated — New York City Local Law 144 and the EU AI Act among others — and how you deploy this in your jurisdiction is a decision to take with your own counsel. We will build to whatever audit trail and record-keeping that requires, and we would rather have that conversation before you buy than after.
What a talent team asks first
Every answer is in the page source rather than behind a click, because the systems that summarise this page never click anything.
How does VectorHire actually evaluate a candidate?
Four agents work on the same candidate at the same time: one reads the résumé, one reads the LinkedIn record, one reads the GitHub history where the role is technical, and one conducts a live voice interview. Each works against a rubric generated from your own job description, and each conclusion arrives attached to the evidence it came from — the specific line, the specific commit, the specific answer.
Does the AI make the hiring decision?
No, and it is not built to. What it produces is a one-page brief per candidate: the scores, the evidence behind each one, and a recommendation. A hiring manager reads it, sees exactly why a candidate scored the way they did, and can overrule any part of it. The decision stays with your team, which is both the design intent and what this category's regulation expects.
Is this fair? What about bias?
We make a process claim rather than an outcome claim, because an outcome claim about fairness is not demonstrable and we will not make one. What the system does is apply the same defined criteria to every application, in the same depth, and show its evidence — where a manual first pass gives different attention and different criteria to each CV depending on who read it and when. Automated employment tools are regulated (New York City Local Law 144, the EU AI Act among others); how you use this in a given jurisdiction is a decision to take with your own counsel, and we will build to whatever audit and record-keeping that requires.
How fast is it?
Fast enough that time-to-first-response stops being the reason you lose people, and the capacity runs to hundreds of candidates an hour. The reason is structural rather than magical: the four agents run in parallel instead of queueing, so a candidate is not waiting for the résumé pass to finish before the interview is offered.
Does it replace our applicant tracking system?
No. Your ATS stays the source of truth and VectorHire writes back into it — candidates, interview results and status. It plugs into the ATS you already run rather than asking you to move, because a screening tool that requires a migration will not survive first contact with a recruiting team.
What about candidate data?
It is collected with the candidate's explicit consent, processed for the assessment, and deleted on request. Public professional data is used as public professional data. We will describe our actual security posture in detail on a call; what we will not do is put a compliance certification on this page that we do not hold.
Can we change the criteria?
That is the point of the rubric being generated from your job description rather than baked into the product. You adjust the weights, add criteria of your own, and see the effect on the evidence rather than on a score alone.
Has a company run this in production?
VectorHire is our own product, and we run it. There is no named client reference in recruiting yet, and we will not invent one. What we offer instead is the honest version: give us a real open role and a batch of real applications, and watch the four agents work them in front of you.
The voice interview is one of three voice systems we run, and the screening workflow can be built into your own stack as an engagement rather than a product.