In modern software engineering, approximately 20-40% of the time is spent on the processes for building, integration, and testing. This staggering number indicates huge challenges in scientific computing software development where many users and developers are not formally trained as software engineers. Today, the scientific computing software stack has become increasingly complex for deriving the full capability of modern high-performance computing systems along with the increasing variety of system environments. For example, the Extreme-scale Scientific Software Development Kit (xSDK) in the U.S. DOE Exascale Computing Project (ECP) depicts a complex set of software dependencies to support the interoperability of 24 scientific library packages. Although the xSDK is one of the special cases, many ECP applications today are using a subset (several packages) of the xSDK suite with a variety of compiler, runtime, and system software combinations, indicating the difficulty of their deployments. Recently, DevOps has become an essential software engineering discipline in enterprise computing; but the uniqueness of scientific computing, including the common use of legacy packages (that have not been tested properly) and the notion of “numerical correctness,” often specific to applications and algorithms (numerical correctness is not understood well in terms of the modern software engineering practice) has made the deployment even more challenging.