Screen the stack. Never touch a cloud with candidate data.
Drop in a resume and the job posting. Get a ranked read back — on this machine, not someone else's server. Recruiters spend up to 23 hours screening resumes for a single hire — this cuts that to minutes, on-device.
Why this room exists
Several competitors in this category (resumescreening.ai, AI CV Ranker, HaiTalent) claim privacy without actually being on-device. This one genuinely never uploads a file. Choose PDF, DOCX, or TXT in — a ranked read out — the same proven engine already shipping in the Paper and Score & Match rooms, not a new, unproven system. Nothing leaves the machine: only a small, structured result is ever shown, the same distillate-only rule already proven in this build.
Screen the whole stack
Rank
Candidate file
Match score
Keyword matches
Contact found
Dates found
Amounts found
Shape check
Matches
How the ranking works. Drop the whole stack in one go. Each resume runs through the same real intents (entities, find, extract, profile) already proven in the Score & Match room, once per file, entirely on this device. The results assemble into one table sorted by match score, highest first — export the whole ranked list as a CSV when you're done.
How it works
1 · Drop the job and the whole stackSelect every resume file at once — all of it stays on this device the whole time, nothing is sent anywhere to be read.
2 · Each file runs the same real intentsEntities pulls contact info; find/extract count keyword matches; profile catches a malformed file before it's scored at all — once per resume, in order.
3 · One ranked table, highest match firstEvery row traces back to a real match in real text — no opaque "AI score." Export the whole list as a CSV when you're done.
Common questions
Does this replace a human recruiter's judgment?No — it replaces the 23-hour first pass. It surfaces real keyword matches and real contact/date/amount entities per candidate so a human's attention goes to the ones worth reading closely.
Is this a black-box AI score?No. Every number shown is a literal count from a real, named intent (a keyword match count, an entity count, a shape check) — nothing here is an opaque model output.
What if the resume is a scanned image, not text?The Shape Check column flags a near-empty extraction so you know to re-request a text-based file before trusting a zero score.