How It Works

Five core stages. Nine intelligence layers.

The core pipeline runs every session from document upload to final verdict. The intelligence layer adds explainability, behavioral signals, trust history, and team-based review on top.

Core Verification Pipeline

Step 01

Upload ID Document

Submit a government-issued ID — passport, national ID, or driver licence. Both front and back are captured where required. The engine validates document type, issuing country, and completeness.

Document type detected
OCR extraction initiated
Validity check started

Step 02

Capture Selfie

A biometric selfie is captured in the session. The engine checks for single-face detection, image quality, and alignment before accepting the capture. Behavioral signals are collected silently.

One face detected
Image quality approved
Behavioral signals collected

Step 03

Run Liveness Check

A liveness challenge confirms the individual is physically present. Active checks (blink, head turn) or passive analysis are applied. Device stability and hold angle are also recorded.

Liveness challenge completed
Passive analysis applied
Live individual confirmed

Step 04

Compare Faces

Multi-source face matching runs between the ID document photo, the captured selfie, and any linked profile photo. Each pair receives an individual score. The face hash is checked against the silent watchlist.

ID Photo ↔ Selfie compared
Watchlist screening complete
Match scores generated

Step 05

Generate Confidence Result

The Identity Consistency Engine aggregates all signals into a single verdict. An Agent Reasoning Panel explains why the score landed where it did. The Trust Score timeline is updated for returning subjects.

Consistency score computed
Agent reasoning generated
Trust score updated

Intelligence Layer — Beyond Ordinary KYC

9 modules

Trust Score Timeline

Every subject builds a living score across repeat verifications — a trust passport that improves with good history.

Behavioral Biometrics

Device hold angle, tap rhythm, and stability are captured silently — invisible to coached subjects, revealing to the engine.

Cross-Context Intelligence

Face hashes are compared across sessions and contexts. Same biometric, different name? The system flags it quietly.

Silent Watchlist

Hash-only first-party screening. No external service, no OFAC overhead, no PII. Your fraud layer, fully private.

Agent Reasoning Panel

Plain-English breakdown of exactly why a result landed where it did — facial landmark delta, OCR confidence, liveness signal strength.

Re-Verification Tokens

Only one step failed? Issue a challenger token. The subject retries that step only — no full session restart.

Smart Profiles

Context-tuned thresholds per use case. Drivers need 88+, general signups pass at 70. One engine, many bars.

Session Replay

Timestamped, tamper-evident recordings of every subject session from consent screen to final result.

Team Collaboration

Async reviewer threads with @mentions, pinned decisions, emoji reactions, and second-opinion requests on every case.

Start a Verification Session