Edge AI for Manufacturing

See everything.
Miss nothing.

Industrial quality control that deploys in 60 seconds. No cloud, no computer, no expertise required.

30 FPS inference
<12ms End-to-end latency
60s Setup time
$0 Cloud cost

Product

Machina Instant Guard

AI-powered visual inspection for high-mix, low-volume manufacturing. Teach the system what "good" looks like in 60 seconds — it handles the rest at 30 FPS.

One-Minute Teach

Hold a button for 3 seconds. The system captures and learns a reference. No datasets, no labeling, no configuration files.

Physical Buttons Only

Green to teach. Red to reset. Designed for operators wearing gloves who never touch a keyboard.

Fully Air-Gapped

All inference runs locally on a Jetson Orin Nano. No WiFi, no cloud account, no data leaves the device.

Real-Time Detection

TensorRT-accelerated inference pipeline with sub-12ms end-to-end latency. Anomalies trigger an instant alert.

Machina Instant Guard v1 — first prototype demo

Open Source

EdgeRunner

Real-time edge AI vision system for industrial safety and autonomous monitoring. Dual-thread architecture combining YOLO detection with VLM reasoning — running on a single Jetson Orin Nano.

edge-runner Apache 2.0

YOLO TensorRT detection at 28+ FPS, VLM inference via Qwen3-VL 2B with constrained JSON output, ByteTrack multi-object tracking, and a live MJPEG streaming dashboard.

28+ FPS
~25W Power
YOLO Detection
VLM Reasoning

Technology

Engineered for the Edge

A zero-copy inference pipeline from camera sensor to GPIO output, built for Jetson Orin Nano with TensorRT acceleration.

Camera GStreamer NVMM
TensorRT MobileNetV3 FP16
PCA 960d → 64d
Scorer Mahalanobis
Output GPIO / Relay
<12ms Total pipeline P99
3–5ms TensorRT inference
$200 Hardware BOM cost
0 External dependencies

Contact

Let's build something.

Looking for pioneer manufacturing partners to deploy Machina Instant Guard. If you run a factory or work in quality engineering, we'd like to talk.