High quality cameras, rich wireless connectivity, and augmented sensors available on smartphones and tablets, combined with their increasing computing and communication capabilities, have enabled many new applications. Mobile augmented reality (MAR) and location based services (LBS) are exemplar categories of such emerging apps. While the extreme integration of a mobile system presents a great opportunity for these new apps, it also imposes an extra burden on app development in order to achieve a satisfactory user experience and energy efficiency for these emerging apps. In this thesis, I address two key challenges of realizing these emerging apps on a battery-powered mobile device. First, mapping a compute-intensive app to a heterogeneous multi-core platform has a huge and complex program optimization space. We address this optimization problem by characterizing the computation and power efficiency of mobile cores and accelerators (e.g., the CPU, GPU, and DSP) and developing solutions to assist the app-to-platform mapping based on analysis of the algorithm's data access patterns. We further develop an indoor localization solution which dynamically combines measurements from multiple sensors to jointly estimate the location of a mobile user in an indoor environment. The low quality sensors embedded in the phone often lead to poor location estimations. To boost the estimation accuracy, we develop an adaptive sensor fusion method which combines location estimations from different sources by adaptively calculating the confident level of each estimation source.