Security AI that stays on-premises.

Face-based access control, perimeter monitoring, intrusion detection, and tailgating prevention — running entirely on-device at the point of capture. No biometric data leaves the building. Designed for GDPR, BIPA, and enterprise privacy compliance from day one.

Four capabilities. One architecture.

NovaEyeD — face-based access control.

Complete on-device facial recognition pipeline — face detection, alignment, feature extraction, and recognition — all running on the access control device. No biometric data transmitted. Works without network connectivity. Enrollment managed via companion mobile app and synchronized fleet-wide.

Perimeter intrusion detection.

Computer vision detection of unauthorized zone entry, running on edge camera hardware at the perimeter. Real-time alert generation on-device. No video stream transmitted for central analysis — only structured alerts exit the camera, routed to your existing security management system via local API.

Safety zone enforcement.

Detect personnel in hazardous or restricted zones and trigger access interlocks, alarms, or machine stop signals on-device. No server round-trip in the safety chain — the response fires from the same device that detects the entry.

Tailgating detection.

Detect multiple individuals entering through a single credential event at controlled access points. Flag tailgating and piggybacking attempts in real time, without dedicated manual monitoring. Builds on people-counting and access-control integration.

Built for biometric compliance — from day one.

NovaEyeD’s on-device architecture isn’t compliance-friendly as a feature. It’s compliance-friendly as the foundation. Biometric feature vectors are computed and stored on the access control device — never centralized, never transmitted, never available for unauthorized retrieval at scale.

GDPR Article 9

Biometric feature vectors processed on-device, not transmitted. Simplifies the Article 9 processing-basis requirements for controllers and processors.

BIPA (Illinois)

On-device processing avoids collection-and-transmission patterns. Reduces consent and disclosure obligations under Illinois biometric privacy law.

UK GDPR & DPA 2018

On-premises processing reduces controller obligations for biometric data transfers under the UK regulatory framework.

Enterprise & government privacy frameworks

Documented architecture and processing notes available for qualified prospects to share with their legal, compliance, and privacy assessment teams.

ModelNova does not provide legal advice. Compliance documentation and architecture notes are available to qualified prospects for review by their legal and compliance teams.

Shipped: NovaEyeD across four silicon platforms.

NovaEyeD is production-deployed on STM32N6, Alif Ensemble, Infineon PSoC Edge E84, and Synaptics Astra. The specifications below are measured from running compiled binaries on the actual silicon — not vendor-supplied estimates.

On STM32N6: ~70ms end-to-end authentication. 96%+ real-world accuracy. 1.7MB SRAM, 2.6MB Flash. Handles peak entry throughput at sub-100ms per authentication without queuing — designed for high-traffic main-entrance deployments.

Total biometric data leaving the device: zero. By architecture, not by setting.

Go deeper.

NovaEyeD

Full product page — technical specs, hardware support, licensing model, and enrollment architecture.

Custom Model Development

Beyond NovaEyeD — bespoke security AI models for your environment, your threat model, your hardware.

ModelNova Zoo

Browse the broader catalog. Perimeter, safety zone, and people-counting models that pair with NovaEyeD for full-stack on-device security.

Talk to security engineering.

Tell us about your facility and your compliance regime. We’ll come back within two business days with a scoping conversation. Compliance and architecture documentation available for your legal team on request.