Digital twins have also made their way into the medical sector in recent years. Medical avatars (MedA), as a further development of digital twins, enable the evaluation of therapeutic approaches or the testing of the effects of various pharmaceutical substances on the individual human organism in silico. Our approach conceptualizes MedA as modular, multi-layered models that integrate individual patient data, biophysical simulations, and AI-based predictions into a unified system. Through interoperable interfaces and a central orchestration layer, data sources, drug models, and simulation-based predictions can be dynamically linked. This creates a flexible, expandable overall system that can be used both for virtual clinical trials and for patient-specific therapy decisions.