This chapter reports on the motivations, design, and challenges of building the Corpus of Chinese Academic Written and Spoken English (CAWSE), an open-access corpus of Chinese students’ English samples collected from a Sino-foreign university in China. To date, the corpus comprises over 1.5 million words in the subcorpus of written assessment, a subcorpus of 63.8 hours of spoken assessment, and a multimodal subcorpus totalling 24 hours of audio/video classroom data, including 107 tasks already or in process of being transcribed and annotated. Focusing on the design and construction of the multimodal subcorpus comprised of video recordings of student interaction in classroom settings, we report specifically on the challenges that arise from the institutional and material realities of working with gatekeepers, handling data ethically, and selecting and processing data that is representative. We also explain the motivations behind focusing on student language, which we characterize as an emerging variety of L2 academic English as a Lingua Franca (ELF). We hope to provide insight into the complexities of developing a multimodal corpus from the unique setting of a foreign-based EMI campus.