The People-Aware Computing group develops machine learning methods for analyzing and interpreting people's context, activities and social networks from mobile sensor data. As sensor equipped mobile devices become commonplace, they can be used to enrich and support communication and collaboration, measure and improve task performance, and transform the assessment of health and wellness via continuous behavioral analysis. A key challenge in this domain is to build efficient machine learning algorithms that transform low-level sensor traces collected in noisy real-world environments into descriptions of high-level human activities. These systems need to learn model of behavior without requiring significant human effort. We have developed new models which parameterize people's behavior in a human-interpretable yet computationally efficient manner, making them more accessible to researchers from other disciplines. The eventual goal of our research is to build systems that recognize individual and group-level activities in a manner which is robust, self-extending, and unobtrusive.