Ambient intelligence is the ability of observing people and their actions even when they are separated from the observer by walls and occlusions, capturing their physiological signals without putting sensors on their bodies, understanding their emotions even when they do not show on their faces, and detecting specific neuro-behavioral diseases before they are clinically diagnosed. We argue that all this information is encoded in radio waves that bounce off people’s bodies, and hence can be captured in a passive and touchless manner. This talk will describe new technologies and machine learning models that analyze radio signals using specialized neural networks to decode and interpret this information, including the systems foundations that allow us to take such technologies from a lab setting to the real-world. The proposed research fundamentally changes what is feasible today and expands our understanding of how to sense and interact with people and objects without physical contact.