Masker - PII Redaction for Medical Documents

At a glance

In short	Upload a medical document and Masker blacks out every piece of personal information on every page, then gives the file back.
Problem	Medical reports must be shared without names, IDs or contact details, and hand redaction is slow and easy to get wrong.
What it does	Finds personal and health information in images and PDFs and covers it with solid black boxes.
How it works	OCR gives the exact position of every word. Gemini reads only the text and decides which words are personal data. Those boxes get blacked out.
Why this design	AI vision models read well but their boxes drift between runs. Letting OCR own the positions makes the redaction land on the exact pixels every time.
Tech	Google Vision OCR, Gemini 2.5 Flash-Lite, Firebase Hosting, Cloud Functions
Live demo	masker-ganit.web.app

Handles JPG, PNG and multi-page PDFs
Detects names, addresses, phone numbers, emails, dates of birth, patient IDs, Aadhaar / SSN, insurance numbers and signatures
Returns the same format it received: PDF in, PDF out
Shows the original and the redacted version side by side