Designing Artificial Intelligence (AI) solutions that can operate in\nreal-world situations is a highly complex task. Deploying such solutions in the\nmedical domain is even more challenging. The promise of using AI to improve\npatient care and reduce cost has encouraged many companies to undertake such\nendeavours. For our team, the goal has been to improve early identification of\ndeteriorating patients in the hospital. Identifying patient deterioration in\nlower acuity wards relies, to a large degree on the attention and intuition of\nclinicians, rather than on the presence of physiological monitoring devices. In\nthese care areas, an automated tool which could continuously observe patients\nand notify the clinical staff of suspected deterioration, would be extremely\nvaluable. In order to develop such an AI-enabled tool, a large collection of\npatient images and audio correlated with corresponding vital signs, past\nmedical history and clinical outcome would be indispensable. To the best of our\nknowledge, no such public or for-pay data set currently exists. This lack of\naudio-visual data led to the decision to conduct exactly such study. The main\ncontributions of this paper are, the description of a protocol for audio-visual\ndata collection study, a cloud-architecture for efficiently processing and\nconsuming such data, and the design of a specific data collection device.\n