Software is a pervasive element of twenty-first century life and an integral element of scientific research. Research in Computer Supported Cooperative Work (CSCW) in recent decades investigates how distributed, collaborative scientific projects take place across different geographical and temporal scales through the enactment of research infrastructures. This dissertation expands upon existing CSCW research with a qualitative, episodic study of a group of cosmologists who are themselves enacting and working among multiple research infrastructures by producing data analysis software as part of a multinational radio telescope project. I describe this cosmology group’s software production practices to explain how software is a material for expressing their scientific method. Software operationalizes and encapsulates their cosmology theory, a model of the telescope, observation data, and ongoing analysis decisions. I demonstrate how by using plots (visualizations of observation data, their software, and the physical telescope) they engage in rigorous and thoughtful testing and analysis of infrastructural components in their work. Doing this data-intensive scientific work requires that they collectively develop a deep understanding of multiple infrastructures to isolate and remove flaws in their data and do a high-precision scientific analysis, interrogating the many embedded relations among conventions of practice that make up their work. My dissertation offers a novel perspective on the production, use, and work of software in science that emphasizes that software in scientific research is not some static product to simply be sustained but a perpetually mutable expression of method to be iterated upon and improved through unfolding research work.