RECRUITMENT SCREENING BY BRAIVEX
Every score quotes its CV.
Screening that shows its working. The model extracts facts. Python does the arithmetic.
- 0 auto-rejects, ever
- Evidence quote per criterion
- Every batch undoable
- Evidence quote per criterion
- Zero auto-rejects
- Deterministic engine
- Undoable batches
- USD billing
Three steps, then humans decide
The software prepares. People choose. That order never flips.
Bring your own AI key
Anthropic, OpenAI, or a free OpenRouter key. Keys stay write-only in Settings.
Drop CVs in a folder
Press Scan folders. The preview names count, folder and rubric before anything is spent.
Open the evidence
Ranked shortlist. Every criterion quotes its CV line. A failed requirement keeps its score.
How it reads
One row of a shortlist. Each number carries the line it came from.
TILL EXPERIENCE · 0.9Operated tills solo on late shifts for two years.
EVENING AVAILABILITY · 1.0Available weekends and late evenings.
Privacy is architecture, not a promise
The safeguards are enforced by the software and its tests — not by policy pages.
Humans decide, always
A failed requirement labels a candidate, never deletes them. The full score and record stay reviewable.
Every number has a quote
Scores come from one deterministic engine; the model supplies facts with verbatim CV lines. Reproducible byte-for-byte.
Nothing looked up, nothing inferred
No social-media searches, no photos, no AI-writing detection, no school or name signals. Gaps and part-time work are never scored.
Roles, invites, write-only keys
Owner, admin, member, viewer. API keys display as set-or-not, never the key. Every provider change is logged.
12 months, then swept
Applications expire on a defined schedule with a one-command sweep. Demo data is fiction and purges itself on first real arrival.
Notices ready to publish
Candidate-facing Art. 13 wording, DPIA, ROPA and balancing test ship as drafts — fill in your controller details before going live.
Every job board. Zero integrations.
LinkedIn, Indeed, Reed, Totaljobs, CV-Library, Monster, ZipRecruiter, Glassdoor, Guardian Jobs, Jobsite — screening starts from the CV file, so every board feeds the same pipeline. No API agreements, no passwords shared, no scrapers.
Download the CVs
Every major board exports applications as files. Drop them in the role folder — Word and photos convert automatically.
Or forward the emails
Application emails land in the inbox folder. Confirm the role per batch and screen — nothing imports silently.
Intake to shortlist, hands-free
Folders, batches, evidence, pipeline stages, retention. The machine prepares; a person approves every rubric and every hire.
Pricing in dollars, honesty included
Limits are displayed, never silently enforced. Canceling keeps your data.
- 2 seats
- 25 screenings / month
- Kick the tyres. 25 screenings, then pick a plan.
- 10 seats
- 500 screenings / month
- The whole hiring team, multiple roles.
Questions, answered plainly
The model reads each CV and extracts facts: a 0–1 factor per criterion, each with an evidence quote. Python turns those facts into numbers. The model never produces a score, so every score reproduces byte-for-byte.
It extracts dated stints, duty levels and evidence quotes. It does not score, rank, or reject. Nothing auto-rejects: a failed requirement keeps its full score, and a person makes every decision.
On your machine, in the project data folder. Retention is 12 months from application, with an automated sweep covering records, CVs, the registry and the event log.
Four USD plans: trial, starter, team, scale. Limits display on the billing page and are never silently enforced. Without a Stripe key, paid controls say so instead of pretending.
Yes. Bring an Anthropic, OpenAI, or free OpenRouter key, pick provider and model in Settings, and screen. Keys stay write-only.
Your first shortlist is ten minutes away
Create your organisation, connect a key, drop in a folder of CVs.