A large-scale, collaborative computational experiment involves more than the tuning of individual code fragments- typically, codes must be assembled, integrated in one or more programming models and languages, experimental runs must be managed and data collected.  In this talk, I will present my work using the Swift programming language to assemble and run many different scientific applications on wide range of computer systems.  In high-performance computing, I will present the Swift/T system for high performance dataflow applications that run on extremely large systems such as the IBM Blue Gene/P and the Cray Blue Waters system.  The techniques presented will cover runtime and compiler development in the Swift context, management of big data for large scale computational experiments, and a variety of application case studies.