Materials science is increasingly turning to data science to extract meaningful inference from large data streams. However, collaborations between researchers within the materials and data-science fields face several challenges including language barriers and domain-specific norms and expectations. In this presentation we discuss our experiences with using a “hackathon” to bridge this gap. We discuss two hackathons models in which participants work intensely for a short time in small groups to develop data-science solutions to authentic materials-science problems. The first hackathon model was implemented in an international workshop with established scientists, while the second model was implemented in a local environment comprised mostly of graduate students. In both cases, we found that short periods of intense interaction resulted in productive interdisciplinary teams. In the talk we discuss the advantages of various formats, approaches that were and were not effective, and offer suggestions for future endeavors and best practices.