Our newly developed edge computing node for AI applications, codenamed “SJAIedge”, has the following salient features: 1) Modular computing and extension units. We used Lattice FPGA (ICE5LP4K) to glue the ST’s Cortex M4 MCU, external ARM/GPU processor (NVIDIA Jetson, Android Things, and Raspberry Pi), multiple radio chips, audio components, sensor modules, and extension ports together. This structure can help the designer select chips and connections for different applications, while keeping the rest of the chips in sleep mode. We support small profile M.2 Key B and Key M as high speed module connectors for Solid-state drive (SSD), external FPGA, and high speed sensors. 2) Powerful on-board stereo camera. Our SJAIedge is equipped with on-board 12-megapixel Sony IMX362 stereo camera modules with HDR support, 4K60Hz, dual-pixel auto-focus (max 24-megapixel output), and large 1.4um pixels for better light sensitivity. We implemented on-board 4-axis optical image stabilization (OIS) and hardware-accelerated image processing via an image signal processor (ISP) and codec inside the NVIDIA Jetson. Our SJAIedge also includes high-power LED flash, ST’s laser ranging sensor (VL53L1X), and two additional CSI 2-lane interface for external camera modules (compatible with Raspberry Pi camera v2). Our on-board camera configuration represents the top-tier customized stereo and 360 degree camera with open configuration and raw pixel data access. For example, Google Pixel 2 only uses one IMX362 camera sensor, and the ZED 3D camera only supports 720p. 3) Flexible and efficient power modules. Our circuit can intelligently select the available power sources (i.e., USB type-C power delivery (PD), power sockets, and multi-cell Li-batteries), perform power negotiation, charge the battery, and support burst current requirement when using old batteries.