Spatially distributed embedded systems that are collections of several computational modules are relatively widespread. Cyber-physical systems (CPS) which tightly integrate traditional embedded computing with the physical space through interactions using sensors and actuators are rapidly gaining significance. Lately, sensors have become an indispensable part of pursuing research in several fields of scientific study, with data gathered using sensors providing the much needed empirical evidence and basis for our principles, models and decisions. Sensor networks have been theoretically studied and practically realized extensively in the last couple of decades. Designing vast networks to sample the physical space is no more a challenge, and the high-performance computing systems (HPC) to assimilate the sensed data are becoming increasingly easy to approach. However, the true potential of combining CPS with HPC is yet to be realized and presents several challenges and opportunities for research in embedded computing and traditional data and processing related activities. My post-doctoral research has focused on understanding this problem space. In this talk I will motivate the need for the strong coupling between CPS and HPC, describe some of the application areas identified and pursued, and discuss future challenges in increasing their cohesion. A path for developing the next-generation of sensor driven HPC platforms and computationally guided feedback in near real-time will be explored. Some of the specifics that will be discussed include hierarchical computing in the network, data aggregation and dissemination approaches, and engineering challenges in realizing a system that can co-exist and evolve over time for performing longitudinal studies.