The sensation of a drink in the throat is a salient example of the internal bodily feelings that shape our eating experiences. Computationally modeling these sensations would enable their redesign and inform technologies that augment how we eat. However, methods for quantifying such subjective, internal states from objective cues remain underdeveloped. This paper introduces a computational approach to bridge this gap. A first study (N = 31) models subjective ratings from laryngeal skin temperature and ingested volume, revealing distinct, individual Interoceptive Profiles. Informed by these findings, we developed a wearable device that provides thermal feedback to the larynx. A second study (N = 20) demonstrates that this intervention can alter drink sensations, contingent on the user’s sensory profile. Based on these findings, we highlight the potential of the larynx as a site for bidirectional interaction (sensing and modulating) and propose a novel approach for personalized sensory augmentation.