EAI EDGE STACK

eAI Edge

Intelligent Edge Node with Neural-Interface Input

One git clone, one cmake, one smoke test. The complete eNI → eIPC → eAI pipeline as a manifest-pinned stack. Deploy a full neural-interface AI node on any EoS device in under 10 minutes.

v0.1.0

Current Version

v0.1

Profile

1024ch

Max Channels

<5mW

Always-On Power

The Pipeline

Three components, one manifest. eNI acquires, eIPC transports, eAI infers.

eNI

Neural / Sensor Input

Acquires neural signals (EEG, EMG, ECoG, LFP, Spikes, fNIRS) and sensor data. Applies hardware-accelerated filtering and spike sorting. Outputs structured EIPC messages.

eIPC

Secure Transport

Routes neural data between eNI and eAI with HMAC-SHA256 integrity, AES-256 encryption. Supports shared memory, UART, SPI, and TCP transports.

eAI

On-Device Inference

Runs manifest-pinned ML models (TFLite, ONNX, GGUF) on the device NPU or CPU. Outputs decoded intent, classifications, or generated text via EIPC.

Quick Start

Deploy the full eAI Edge stack in 3 commands.

# Clone the eAI Edge stack

$ git clone https://github.com/embeddedos-org/eAI

$ cmake -B build -DPLATFORM=rk3588s -DMANIFEST=manifest.yml

$ cmake --build build && ebuild flash

# Smoke test: verify all 3 pipeline stages are running

$ ebuild monitor --filter eai-edge

✓ eNI: 64ch EEG @ 1kHz — OK

✓ eIPC: 0.4ms transport latency — OK

✓ eAI: gesture_classifier_v2.tflite loaded — OK

Use Cases

Motor BCI Prosthetics

Decode motor intent from 64-channel ECoG → eIPC → eAI decoder → robotic arm control.

Seizure Detection & Suppression

Continuous EEG monitoring → eAI seizure classifier → eIPC → neurostimulator trigger. <2s detection latency.

Gesture Recognition

8-channel EMG → eNI spike sorting → eIPC → eAI gesture classifier → EIPC actuator command. 50ms latency.

Cognitive Load Monitoring

Frontal EEG → eAI fatigue model → EIPC alert → eOffice notification. Always-on at <5mW.

EAI Platform Neural Link AI