As part of the recent news that was announced at Embedded World 2025 (spoiler: we've been acquired by Qualcomm), the Edge Impulse team released our first supported Qualcomm device on the platform, the RB3 Gen 2 Developer Kit. This is a full-featured robotics, automotive, and computer vision platform, aimed at professional developers building enterprise projects. However, there are other boards available from other Qualcomm partners, targeting different markets and price points. One of those is the Thundercomm Rubik Pi 3, which uses the same Dragonwing 6490 SoC found in the RB3 Gen 2, but in a more traditional single board computer form-factor.
Like the RB3 Gen 2, the Rubik Pi also runs Qualcomm Linux as its native OS. In order to test out the process of running computer vision models on the Rubik Pi, we simply followed the same instructions we had written for the RB3 Gen 2, and sure enough the Hexagon NPU on the Rubik Pi was able to load an Edge Impulse model and run inference. Here's a quick video describing the board and its features, as well as an overview of the steps needed to run an accelerated edge AI model on the device:
If you purchase a Rubik Pi 3 and build a project with it, be sure to let us know over in the Edge Impulse Forums, or be sure to contact us with any questions!