Abstract
This document presents the product briefing and the product requirements document (PRD) for a 16- week pilot of a virtual patient designed to train medical students' clinical reasoning and to adjust case difficulty to each student's academic term. The patient simulates symptoms, tests, habits and reactions to treatment, and responds to the student's decisions from diagnosis through follow-up. A rule-based clinical engine controls the medical content and guarantees clinical consistency, while a large language model gives the patient a voice and leads the post-simulation debrief. The pilot covers one base case (chest pain in the emergency department) at three difficulty levels, with about 90 students and an equivalent comparison group. The document defines the problem, personas, scope, hypotheses, success metrics, risks and governance, 20 functional requirements, a difficulty model by academic term, a competency rubric, non-functional requirements, a high-level architecture, a pilot plan with two decision gates, and the acceptance and go/no-go criteria. All figures and deadlines are proposals to be calibrated with the institution's course coordination.