PROJECT SUMMARY / ABSTRACT The vast majority of pediatric and adult cardiac surgery procedures worldwide involve the use of cardiopulmonary bypass with a dedicated and highly customized perfusion system. Cardiac surgery procedures are complex and entail safety-critical activities, requiring continuous coordination between four subteams: perfusion, surgical, nursing and anesthesia. This environment demands outstanding technical and non-technical skills (e.g. teamwork, communication and situational awareness). Over the past few decades, technological advancements have improved the safety and efficiency of the perfusion system, however, despite substantial progress, recent studies continue to report a high incidence of preventable intraoperative adverse events among cardiac surgery patients. The perfusion system, in particular, relies heavily on the expertise and skills of the perfusionist, and there is currently no computational intelligent system to support perfusionists' optimal decision-making during the critical phase of cardiopulmonary bypass. In this proposal, we seek to develop a data-driven approach to learn from expert perfusionists how to achieve optimal outcomes for cardiac surgery patients. Rather than attempt to engineer a solution, we propose to develop a computer-based apprentice that can learn from high-quality demonstrations of perfusionist actions to infer gold-standard patient care. Our goal is to develop and evaluate a Robot-Assisted Perfusion System (RAPS) that can be integrated into the cardiac surgery workflow as a non-human teammate. The RAPS will support the perfusion team in a way that perfusionists still will keep control of the perfusion system (i.e. human- in-the-loop), but cognitively supported and guided by the RAPS.