The MIC-743-AT utilizes a full sensor-to-inference pipeline, from multi-format camera input to real-time AI processing at the edge, all connected through high-speed ethernet with minimal latency. The system integration begins at the sensor level, supporting a variety of camera interfaces such as MIPI CSI and other high-speed industrial protocols.

Sensor data is routed through NVIDIA Holoscan Sensor Bridge, a sensor-to-ethernet streaming platform designed for low-latency applications, and aggregated via Advantech's high-speed sensor switch, which supports up to 2×10GbE uplinks depending on the system configuration. This data is then streamed directly into the MIC-743-AT, a compact, rugged, high-performance AI computing system built on the NVIDIA Jetson Thor module and powered by the NVIDIA Holoscan platform.

Leveraging camera over ethernet and high-throughput networking, the architecture ensures ultra-low latency and efficient data processing, which is crucial for safety-critical mobile and edge AI applications. The MIC-743-AT platform delivers an optimized, compact, and rugged solution for intelligent mobility, VLM (Vision Language Model) at the edge, and humanoid robotics.