Cloud computing holds promise to deliver computing as a utility to scientists and large user communities. While vast computational resources could be made available instantly on cloud computing platforms, system installation and application configuration on top of these resources is still a time- and labor-consuming process, especially for compute- and data-intensive HPC applications running in large clusters. We are going to develop a rapid application provisioning tool that allows users to automate system installation, application compilation/deployment/execution on ephemeral and elastic cloud servers. Using our deliverables, users can also take advantage of cloud-based massively-scalable parallelization framework to process user-supplied (scientific) datasets. We are going to demonstrate our prototype on both public and private clouds using applications, benchmarks and data formats of NASA interest.