Data-driven multi-sensor technology for AI-assisted chemical sensing

The quest to explore the properties of complex liquids beyond the capabilities of conventional analytical chemistry has driven the development of electronic tongue systems. Inspired by the human sense of taste, researchers have achieved unprecedented success in addressing intricate sensing challenges. Yet, much of the focus has been on applicationoriented systems designed to meet specific needs through the tailored combination of selective sensors. The advent of cross-sensitive sensor arrays has drastically shifted this paradigm. Recent advances in materials science, electronics, and machine learning methods hold promise for the seamless integration of low-selective, highly cross-sensitive systems in electronic tongue technologies. This dissertation provides novel insights into the infusion of intelligence into such potentiometric chemical sensor arrays, unlocking unexplored methods and functionalities that enhance their potential for deployment outside of laboratory settings. Specifically, data-driven methodologies are proposed and applied to a model system consisting of a miniaturized sensor array with minimal hardware components, leading to a portable electronic tongue prototype. The reconfigurability, portability, and versatility of this system are demonstrated across various use cases, including the quantification of major constituents in complex liquids, the identification of commercial liquid products, the estimation of human sensory perception of beverages, and the detection of adulterated food samples. The results highlight how portable electronic tongues, supported by advanced data-driven approaches, accelerate the chemical analysis of complex liquids by teaching sensors how to "taste." Additionally, recent advances in machine learning and artificial intelligence (AI) are applied to the field of chemical sensing, demonstrating how sensor array design can be informed by data rather than theoretical models. AI foundation models are shown to enhance sensor performance through transfer learning in data-scarce scenarios. Ultimately, by integrating these concepts into a portable sensor development kit, this research promotes the democratization of electronic tongue technology, bringing it closer to end-users for personalized chemical sensing applications.