Since robots were first introduced into the classroom by Seymour Papert, today, there are two main approaches in Educational Robotics (ER): the adoption of ready-to-use robots or hands-on robotics activities built on Piaget's constructivist theories, mainly applied in STEM activities. The problem is that most robotic kits on the market are based on closed software and hardware, with pre-defined instructions for building and coding robots. This approach does not sufficiently nurture students' creativity, divergent thinking, and innovation. The present study presents a series of STEAM methodologies to create personalized educational robots built by assembling recycled material and everyday objects embedded with Machine Learning solutions. Data have been collected during some STEAM workshops for teachers and PhD students through questionnaires, pictures, and video recording. Results show that do-it-yourself (DIY) practices can successfully integrate didactic content in the robots' design and storytelling, for example, with disciplines such as Mathematics, Geometry, Art, and Philosophy, creating unique and personalized approaches in ER.

Keywords: Educational Robotics, Machine Learning, Coding, STEAM, DIY.