What does Veridexa do?

Veridexa is an enterprise AI platform that verifies documents and detects fraud in identities, invoices, and KYC files using forensic analysis, OCR, and machine learning.

How does Veridexa protect data?

Documents are processed with encryption in transit and at rest on secure cloud infrastructure, with regional data residency options in EU, US, and APAC.

Banks, insurers, fintechs, and marketplaces use Veridexa to automate KYC, catch document forgery, and reduce manual review.

Face Verification

Server-side YuNet + SFace 1:1 face matching (ONNX Runtime Node).

Live Engine

Run Face Verification View API Documentation

Engine

src/lib/verification/biometric-server/face (YuNet detector + SFace embedder, ONNX Runtime Node)

API

POST /api/public/biometrics/v1/enroll · /compare

Face Verification is a full 1:1 engine. On the production API path, YuNet detects a single face, 5-point landmarks drive a similarity transform for aligned SFace inference, and 128-dim embeddings are compared by cosine similarity. Templates are stored AES-256-GCM encrypted with versioned key rotation. A separate in-browser embedder module exists in the codebase but ships with no default model (fail-closed) until an operator supplies one they hold commercial rights to; it is not used by the public API today.

What is implemented

YuNet ONNX face detector
SFace embeddings (128-d)
Quality gate: brightness, sharpness, pose, face count
Encrypted storage in face_recognition_templates
Cosine similarity + decision threshold

Modules

src/lib/verification/biometric-server/face/
  yunet-detector.server.ts
  sface-embedder.server.ts
  pipeline.server.ts
  image-ops.server.ts

// In-browser embedder (disabled by default, no shipped model):
src/lib/verification/biometric-core/face-recognition/
  arcface-embedder.ts

Biometric APIs SDK Back to overview