The medical interview to obtain a patient’s history is a powerful tool that contributes to overall patient health outcomes. In recent years, medical schools have been including the use of virtual patients (VPs) as part of their medical interviewing and interpersonal skills training curriculum. These skills, though, are difficult to teach and learn, especially in under-resourced contexts. Challenges faced include lack of direct supervision especially in a large group patient interview, missing important parts of the history or jumping from one part of the medical history to another, failure to consider cultural context and variable metrics for assessing medical interview skills. In this paper, we describe Caladrius, a VP designed for Philippine medical education and the feedback from initial user tests with medical school faculty and students. The faculty and students observed that Caladrius’ responses need further elaboration; otherwise, the interview becomes unnatural and tedious. The text-based interface performed quickly enough to be usable, that the voice-based interface was too slow, and that the understanding of Filipino, especially health-related Filipino terms was limited. Caladrius is a case study that illustrates the need for low-bandwidth AI solutions trained on specific, curated language subsets to maximize access.