Aadhaar cards
Front and back of the card, PVC or paper, scanned or photographed, including the long letter-style card.
See it work
Real documents are messy. The masking still lands.
Pick a sample to see what the AI detects, and what comes back.
AI-powered Aadhaar masking
Find every Aadhaar number. Mask it before anyone sees it.
Our AI spots Aadhaar cards, letters and e-Aadhaar inside any scan or photo, even blurred, skewed or half in frame, and hides the first 8 digits automatically.
- Your documentimage · TIFF · PDF
- Detectcard · letter · e-Aadhaar · QR
- Maskfirst 8 digits hidden
- Reportstatus for every page
12 crore+documents masked in production
8 of 12digits masked, last 4 kept
4 formatsJPG, PNG, TIFF and PDF, multi-page
On-premisesdocuments never leave your network
DetectedAADHAAR_FRONT · UID
Masked outputMASKED
DetectedEAADHAAR · UID · QRCODE
Masked outputMASKED · 2 numbers + QR
DetectedOCR_UID
Masked outputMASKED
What it finds
Every way an Aadhaar number turns up in your files.
A vision model trained on Indian KYC documents, backed by OCR for numbers that appear as plain text.
Letters and e-Aadhaar
Full Aadhaar letters and downloaded e-Aadhaar PDFs, with every number on the page masked, not just the first one.
Numbers written as text
OCR reads Aadhaar numbers typed or written in forms and letters, and confirms each with the Verhoeff checksum before masking.
Secure QR, when you want it
The QR code carries the full number. Mask it too with a single request flag.
Masked Aadhaar, recognised
Documents that are already masked are identified and reported, so they are never counted as a miss.
Insurance and banking forms
Trained on claim and proposal forms with an Aadhaar field, so numbers on your own paperwork are caught too.
Built for real-world scans
Bad photos are the norm, not the exception.
- Poor qualityLow-resolution, blurred, faded and over-compressed scans.
- Skewed or rotatedCards shot at an angle, rotated pages and sideways phone photos.
- Partly visibleCards cut off at the edge, overlapped or folded.
- Many per pageSeveral cards or numbers on one page, all masked together.
- Mixed bundlesLong PDFs where Aadhaar sits among PAN, passports and other pages.
How it masks
Masked the way the regulations expect.
The first 8 digits are hidden, the last 4 stay readable, and the file comes back in the format you sent.
XXXXXXXX9012
Masked: first 8 digitsVisible: last 4 digits
Removed from PDFs, not just covered
In PDFs, the content under the mask is erased before the block is drawn, so the number can't be copied out or recovered.
Drop-in output
JPG stays JPG, multi-page TIFF stays TIFF, PDF stays PDF, with page order and layout kept as they were.
Smaller files, not bigger
Heavy scanned PDFs and TIFFs are recompressed as they are masked, so your storage doesn't grow.
A status for every document
Know exactly what happened, page by page.
Each response lists what was found on every page, so you can route exceptions to a person.
- MASKED
- Aadhaar found and masked.
- ALREADY_MASKED
- The document was already masked.
- LOW_SCORE_AADHAAR
- An Aadhaar was seen but needs review. Send it to a checker.
- NON_AADHAAR_ID_DOCUMENT
- Another ID document, with no Aadhaar on it.
- NON_ID_DOCUMENT
- No identity document found.
Two ways to use it
Clean up the archive. Then keep it clean.
Mask what you already hold in bulk, and mask new documents the moment they arrive.
For the documents you already have
Point the utility at your existing KYC archive and let it work through millions of files, unattended.
- Large archives: images, multi-page TIFFs and PDFs, processed in batches.
- Parallel pages: pages of each document are processed at the same time.
- Full report: a status for every file and page, so exceptions are easy to find.
- Compliance clean-up: bring legacy records in line with Aadhaar masking rules.
For every new document, as it arrives
Call the API from your onboarding, loan, claims or DMS application and store only the masked copy.
- REST API: send a document, get the masked file and a page report back.
- Client SDK: a few lines of code, with encryption handled for you.
- Async with callback: queue large files and get the result pushed back.
- Your rules: turn QR masking and OCR on or off per request.
Real-time API: how it works
Send
Your app sends the document as soon as it is uploaded.
Mask
The AI finds every Aadhaar on every page and masks it.
Store
You save the masked file and act on the page report.
Simple response
One call. A masked file and a clear answer.
The masked document comes back with a status for the whole file and for each page, ready for your workflow rules.
JPGPNGTIFF · multi-pagePDF · multi-page
{
"status": "MASKED",
"dataType": "pdf",
"maskedData": "JVBERi0xLjcK…", // masked file
"pages": [
{ "page": 1, "pageStatus": "MASKED",
"pageObjects": [
{ "name": "AADHAAR_FRONT", "count": 1 },
{ "name": "UID", "count": 1 } ] },
{ "page": 2, "pageStatus": "NON_AADHAAR_ID_DOCUMENT" }
]
}Security
Built to handle the most sensitive number in India.
Runs in your own data centre, encrypts every call, and keeps nothing it doesn't need.
Your network, your servers
Deployed in your data centre on GPU servers. Documents never go to a third-party cloud.
Encrypted end to end
Requests and responses are encrypted with each client's own certificate, on top of TLS.
No replayed requests
Every call carries a unique ID and a timestamp, and requests that are reused or stale are rejected.
Nothing kept by default
Documents are processed in memory and returned. Audit logging can be switched on when you need it.
Every application, separately
Each calling application gets its own credentials and certificate, so access can be granted and revoked individually.
Easy to operate
Documented error codes and request timeouts let your team handle every outcome without guesswork.