How might we enable engineers without a background in AI to create, test, and interact with artificially intelligent models in the physical world? Project Moab is a hardware kit designed to onboard users to the power of machine teaching by bringing a tangible experience right to their desk. Moab, a ball-balancing robot, leverages computer vision and various control methods to balance objects. Users can select classical control systems such as PID loops, trained AI control systems, or manual control via an integrated joystick. An undermounted camera tracks objects placed on the clear balance plate, which is then dynamically oriented by 3 servo-arm linkages. The robot's reactions and 'skill' depends on the control method, which users can test and evaluate against their own trained models. The hardware kit pairs with a web-based machine teaching service, where users create and train their own AI models or 'brains'. Open-source code and 3D part files are ready for download via QR codes on the packaging and documentation. Moab was designed from the ground up to inspire engineers to think beyond object balancing and solve their own novel use cases via intelligent control systems.