TalentFlow
An AI recruiting portal: a CV goes in as a PDF and comes out scored against the role, with the arithmetic on show.
The walkthrough
From a PDF to a scored candidate62s · No sound
The requirement carrying 50% of the weight is the one she does not meet, and the final breakdown shows exactly why the score lands at 55.
Semantic search13s · No sound
Describe the profile in plain language. The vectors are built without name, email or phone: it searches by profile, never by person.
Abstract
The recruiter splits 100 points across the requirements of a role. The model scores each requirement on its own and the final number is computed in code as Σ(score × weight) ÷ 100 — the model never picks it. That is what makes the breakdown auditable: every score carries the line from the CV that earned it. The applicant, meanwhile, only ever sees received, selected or rejected; the score never reaches their side, and a restricted database view enforces that rather than the front end.
Want this running on your data?
We start by mapping one of your processes end to end, at no cost. The blueprint comes out of that.