University-industry (U-I) collaboration takes on many forms, from research services, teaching and training, to curiosity-led research. In the chemical industries, academic chemists generate new knowledge, address novel problems faced by industry, and train the future workforce in cutting-edge methods. In this study, we examine the dynamic structures of collaborative research contracts and grants between academic and industry partners over a 5-year period within a research-intensive Australian university. We reconstruct internal contract data provided by a university research office as records of its collaborations into a complex relational database that links researchers to research projects. We then structure this complex relational data as a two-mode network of researcher-project collaborations for utilisation with Social Network Analysis (SNA)-a relational methodology ideally suited to relational data. Specifically, we use a stochastic actor-oriented model (SAOM), a statistical network model for longitudinal two-mode network data. Although the dataset is complicated, we manage to replicate it exactly using a very parsimonious and relatable network model. Results indicate that as academics gain experience, they become more involved in direct research contracts with industry, and in research projects more generally. Further, more senior academics are involved in projects involving both industry partners and other academic partners of any level. While more experienced academics are also less likely to repeat collaborations with the same colleagues, there is a more general tendency in these collaborations, regardless of academic seniority or industry engagement, for prior collaborations to predict future collaborations. We discuss implications for industry and academics.