Hardware + silicon + edge platforms

The technology that
sees, thinks, and decides

eye3.ai builds the silicon, camera modules, and edge platforms that turn raw pixels into validated actions — at the edge, in real time, with full sensor awareness.

RK1828 platform

Edge AI processing
built for vision

Powered by Rockchip RK3588 / RK3688 + RK1828. Designed for real-time perception and reasoning at the edge — not retrofitted from generic compute.

  • Multi-camera input — simultaneous streams, zero latency
  • Hardware-accelerated ISP + AI pipeline
  • Local inference — no cloud, no round-trips
  • OEM-ready reference design
RK1828 edge AI board
Developer platform

Built for developers
and OEM integration

Compatible with Raspberry Pi 5 and ARM platforms. Prototype in hours, integrate into product in weeks.

  • Plug-and-play camera modules — connects in minutes
  • Python SDK, REST API, WebSocket streams
  • Hardware-agnostic VAOS — same code, any board
  • OEM licensing for custom silicon variants
Read the docs→
Raspberry Pi with eye3 camera module
Modular vision stack

Every layer purpose-built

Swap, extend, or replace any module. Hardware-neutral by design.

Modular vision stack: Pi base + RK1820 AI module + camera
📷

Camera module

Real-world perception. Wide-angle, low-light, IR, multi-spectral options. Integrates with eye3 ISP.

🧠

RK1820 / AI module

Edge intelligence. Transformer-class inference, hardware ISP, sensor reliability scoring on-die.

🔌

Pi / ARM base

Control + integration. Linux SoC, networking, storage, GPIO, integration with site systems.

Innovation

Three patent-pending
core innovations

🎛️

Sensor-aware perception

Continuous evaluation of sensor reliability — calibration drift, occlusion, exposure, thermal — feeding directly into inference confidence. Patent pending.

🗺️

Scene Contract

Structured, machine-readable perception output that bridges vision models and action systems. Open schema; patent on the validation envelope.

🤖

Validation-gated action

Closed-loop runtime that validates every recommended action against current scene confidence before execution. The trust layer for autonomy.

Hardware support

Runs where you deploy

OEM & licensing programs→

From silicon to ecosystem.

Custom hardware, OEM integrations, or co-design — we work with semiconductor partners, integrators, and product teams.

Hardware enquiries→ Build with VAOS