Platform — Identity
Every user,proved
State-of-the-art identity verification and real-time fraud detection, built for fintechs: the document read and authenticated down to its chip, the face proved live and matched, and every fraud signal scored while the user is still in the flow.
01 — Know Your Customer
01 — MRZ read
The lines a forger has to typeset perfectly.
The machine-readable zone — three lines on the back of a national ID, two at the foot of a passport's photo page — is read and checksummed character by character. Alter one field and the arithmetic stops agreeing with the line it is printed on.
Defeats: reprinted data pages · altered dates · invented numbers
02 — OCR capture
Every printed field, lifted off the page.
Optical character recognition reads the whole document — the visual zone a human sees, not just the machine lines — and returns it as structured fields. What the page prints, what the MRZ encodes and what the chip holds have to be the same person.
Defeats: edited fields · font substitution · zone mismatches
03 — NFC chip
A fake can copy the plastic. Not the key.
Biometric documents carry a chip signed by the state that issued them. Lay the phone on the document — near field means near — and the chip answers a live cryptographic challenge a clone cannot fake. The strongest document check that exists, in about a second.
Defeats: counterfeit documents · cloned chips · photo swaps
04 — Liveness
Present, alive — and not a picture of one.
Passive liveness reads a single capture for the texture, depth and motion of an actual face in an actual room — no head-turning gymnastics for the user. Screens, prints, masks and injected video all fail the same quiet exam.
Defeats: printed photos · screen replays · masks · injected video
05 — Face match
The face on the chip is the face in the room.
The live face that passed liveness is matched against the portrait read from the document and its chip. It closes the loop: the document is genuine, the person is real, and — the part fraud actually hinges on — they are the same person.
Defeats: borrowed documents · stolen identities · lookalike fraud
02 — Real-time fraud detection
Caught in the act, not in the audit.
Fraud signals are not a report that arrives after the money has moved. They are scored during the verification itself, and the verdict carries them — so the decision happens while the fraudster is still holding the camera.
Document forensics
Tamper detection on every capture — replaced portraits, edited text, print artefacts and screen recaptures, judged against how the genuine document is actually built.
Injection defence
Virtual cameras, emulators and replayed streams are caught at the capture layer: the feed itself is authenticated before a single frame is trusted.
Device intelligence
The device tells its own story. Rooted handsets, emulators, tampered clocks and repeated attempts all score against the session they come from.
Velocity and reuse
The same face or the same document surfacing under different names is caught across your whole user base — not just inside one session.
Watchlist screening
Sanctions, politically exposed persons and adverse media at onboarding, then re-screened on whatever schedule your policy sets.
Your rules, in the loop
Every signal streams into your policy engine: approve, refuse or route to review in the same call, with the evidence attached to the verdict.
Signal stream
- document.forensicsscreen recaptureREFUSED
- nfc.authsignature verifiedCLEARED
- livenesssingle capture, liveCLEARED
- velocityface seen under another nameREFUSED
- ocr.crosscheckzones agreeCLEARED
- device.intelemulator detectedREFUSED
- watchlistno matchCLEARED
- injection.defencevirtual camera feedREFUSED
- mrz.checksumdigits agreeCLEARED
- face.matchportrait and person agreeCLEARED
03 — How it runs
From a camera to a verdict.
Four movements, and the last one is the point: identity on Pixely is proved once, then proved again — without asking twice.
- 01
Capture
The kit guides the shot — document, face, glare and focus handled on the device before anything is sent.
- 02
Verify
All five checks and the fraud signals run in one pass — print, machine lines, chip, face and device judged together.
- 03
Attest
The result is written to the ledger as a signed attestation — who was proved, when, and by which checks.
- 04
Reuse
The next product asks the attestation, not the person. Onboarding stops repeating itself.
04 — Beyond the document
Verification is the front door.
The platform keeps working after it closes — on the companies behind the accounts, on questions that deserve answers instead of copies, and inside your own stack.
Business verification
Registry records, beneficial owners and signatories for the companies you onboard, resolved to the actual people behind them — the same checks, aimed at a business.
Attribute proofs
Answer questions without copying data. Over eighteen, resident, account holder — attested by the platform, carried by the user.
APIs and SDKs
A REST API and drop-in capture kits for web and mobile, so the first verification is an afternoon's work rather than a quarter's.
05 — Get started
Prove yourfirst user
Create an account and run a verification against the sandbox, or talk to an engineer about your documents and your rules.