Abstract: Under the context of integration between the New Engineering Education and the New Medical Education, interdisciplinary medical-engineering practice courses play a critical role in cultivating students’ capacity for systematic analysis and for solving complex engineering problems. This paper takes the magnetically controlled biopsy capsule robot project in the course Exploration of Biomedical Mechanical Engineering as a vehicle to explore a project-based practice-oriented teaching reform empowered by artificial intelligence (AI). Driven by authentic clinical scenarios, the course embeds lesion image recognition as a key functional module into a closed-loop “recognition-control-sampling triggering” pipeline, and appropriately integrates generative AI tools into literature review, solution justification, code development and engineering documentation. Course quality is ensured through process logging and milestone-based evaluations. Teaching practice indicates that this model helps strengthen students’ end-to-end closed-loop competence from requirement analysis to prototype validation, enhances interdisciplinary collaboration and engineering rigor, and offers actionable experience for reforming interdisciplinary medical-engineering practice courses in the era of artificial intelligence.