Ensuring patient privacy in the use of personal data within biomedical research datasets remains one of the foremost challenges in the field, especially concerning the risk of re-identification from clinical data. This paper presents a conceptual framework for an architecture designed to address this challenge. The model is conceived to, first, allow medical institutions to access clinical data from patient-controlled personal repositories based on patient consent and, second, provide tools for healthcare providers to integrate this data to build larger, more complex datasets. To conceptualize this system, we leverage semantic technologies like Shape Expressions (ShEx), alongside standards such as FHIR and SNOMED CT. The proposed framework integrates these components under SOLID principles and incorporates an LLM layer for accessibility. This theoretical architecture provides a foundation for a decentralized, secure, and accessible system for healthcare professionals.