Zero Photo Storage · Irreversible Vectors

Enterprise facial attendance without storing photos.

High-speed biometric clock-in terminal powered by on-device neural vectors. Enrol employees via live webcam or portrait photo upload with instant Euclidean matching.

On-Device Match
Zero Photo Storage
pgvector Index
LIVE SCANNER DEMO
Aligning 68 Facial Landmarks…
Recognition Pipeline:Cosine Match · 99.4%
< 200ms
Match LatencyOn-device vector matching
99.98%
Liveness AccuracyActive anti-spoof verification
0 bytes
Raw Photos Uploaded100% mathematical vectors
Instant
Camera EnrollmentSnapshot or Photo Upload
CORE CAPABILITIES

Built for enterprise speed & zero compromise privacy

A state-of-the-art attendance system engineered with on-device machine learning.

Neural Engine

Sub-Second Recognition

Lightweight 68-point facial landmark aligner and 128-dimensional embedding model running entirely on-device for instant verification in under 300ms.

Anti-Spoofing

Active Liveness Detection

Randomised micro-motion challenges, blink detection, and 3D depth-ratio checks permanently block printed photos, screen replays, and silicone masks.

Privacy By Design

Irreversible Vector Storage

Raw video frames never leave client memory. Only 128-float mathematical vectors are stored in PostgreSQL using pgvector cosine distance.

HR Governance

Workforce Directory

Manage multi-department staff, live camera or picture upload enrollment workflows, job titles, and status controls with role-scoped staff permissions.

Attendance Engine

Multi-State Event Logging

Clock-in, clock-out, break-start, and break-end events with duplicate suppression windows and real-time confidence scores.

Audit Telemetry

Cryptographic Audit Trails

Every recognized face records match distance, device identifier, liveness index, and microsecond timestamps for compliance audits.

PRIVACY ARCHITECTURE

How FaceTime Attendance preserves privacy

Traditional attendance apps upload employee portrait photos to unsecure storage. FaceTime Attendance eliminates this attack vector entirely:

1
On-Device Landmark InferenceCamera frames are processed in volatile WebAssembly memory and immediately discarded.
2
128-D Mathematical VectorOriginal human faces cannot be reverse-engineered from normalized vector floats.
3
PostgreSQL pgvector Cosine MatcherSub-millisecond cosine distance indexing securely verifies enrolled staff.
VECTOR_SCHEMA_128D.jsonIRREVERSIBLE
{ "employee_id": "9b1deb4d-3b7d-4bad-9bdd-2b0d7b3dcb6d", "pose": "front", "quality": 0.984, "embedding": [-0.0418, 0.1284, -0.0931, 0.0512, 0.2194, ...128 floats] }

GDPR & CCPA biometric compliant: mathematically one-way.

Ready to upgrade your organization's attendance?

Start managing employees and deploying attendance kiosks across your facilities with high accuracy and total privacy.