The unique Physical Reservoir Computing project offers an innovative, energy-efficient machine learning method by using physical systems like a bucket of water. While artificial neural networks (ANNs) have advanced many fields, they require substantial energy, especially for dynamic tasks handled by recurrent neural networks (RNNs). Reservoir Computing (RC) addresses this by transforming time-dependent inputs into high-dimensional patterns using compliant systems, such as water surfaces, which can temporarily store environmental information due to their fading-memory feature.  RC has been applied to various tasks, including temporal pattern classification, time series forecasting, and adaptive filtering, providing a sustainable alternative to traditional deep learning. Physical RC represents a shift towards eco-friendly AI, suitable for edge computing and natural intelligence, aligning with the green technology goals of BUA and TU Berlin.  Participants in this course developed a Physical Reservoir Computer using a bucket of water, basic electronics, and computer vision. Servo motors create ripples on the water as input signals, and a camera captures the surface's response, processed by a Raspberry Pi for low energy use, demonstrating the potential of sustainable AI solutions.