Blind Hire Kit

Four things, each with a conservative, high-confidence rule: the candidate's name (the largest short text block near the top of page 1, or the first short prominent line for Word documents); an embedded photo near the top of page 1 (PDF only - Word files never carry any embedded image into the output at all, regardless of position); a street address (a house-number-plus-street-name pattern, e.g. "1500 Elm Street") found near the top of the document; and a 4-digit year that sits next to a graduation/degree keyword like "graduated", "B.A.", "B.S.", "Bachelor", "University" or "Class of". A work-history date range with no degree keyword nearby (e.g. "2019-2023, Software Engineer") is left alone.

No - this is a bias-reduction aid, not a certification. It's deliberately conservative: it's built to only redact when it's confident, which means it can miss things (a name in an unusual format, a photo placed lower on the page, an address written without a recognizable street suffix, other identifying details like a personal website or a graduation year written out as "2004" mentioned mid-paragraph rather than next to a degree). The hiring panel is still responsible for a final review of each redacted resume before it's used to make decisions.

No. Every output PDF is built from an empty document, and only the text and images that should survive are ever written into it - flagged names, address text, graduation years and photos are simply never included in the first place. The black box you see is drawn purely as a visual marker of what's missing; it isn't painted on top of the original content, because the original content was never copied into the new file to begin with. That was verified directly during development by re-extracting text from redacted output files and confirming the flagged strings were genuinely gone, not just visually covered.

Not exactly, by design. PDF resumes keep their original layout, fonts and page order - only the flagged name/photo/address/year are removed. Word (.docx) resumes have no lightweight way to be rendered pixel-for-pixel as a PDF in a browser, so they're re-typeset into one simple, consistent template with a fixed Experience → Education → Skills order. Either way, every candidate's output is a plain, standard PDF.

No. Reading each resume, detecting what to redact, and rebuilding the output PDF all happen locally in your browser using pdf.js and pdf-lib. Nothing about the files - or the candidates' information in them - is ever sent anywhere.

Up to 20 per batch, processed one at a time with a progress indicator. Each one can be downloaded individually, or you can download the whole batch as one .zip. Output files are named "candidate-1-redacted.pdf", "candidate-2-redacted.pdf" and so on - deliberately not based on the original filename, since a filename like "John_Smith_Resume.pdf" would leak the redacted name right back out.

You'll see a warning on that file. Detection depends on a real, selectable text layer (via the same PDF text-extraction this site's other PDF tools use) - a scanned photo of a resume has no text to search, so automatic redaction can't run on it, and it should be reviewed manually before use.