<p>Deep learning (DL) has appeared as a transformational technology in the medical imaging domain, offering powerful capabilities for complex diagnostic tasks. DL-based models, in particular, have greatly aided skin cancer diagnosis due to their capacity to understand complex patterns from large-scale image datasets. One of the major problems about DL models is their black-box character, which makes their decision-making processes difficult for doctors to understand or trust. This lack of transparency and concerns over the privacy of sensitive medical data present significant obstacles for deploying AI solutions in real-world healthcare environments. To handle the privacy constraints connected with centralized data training, federated learning (FL) has been introduced as a decentralized approach enabling organizations to train models collaboratively without transferring patient’s raw data. Although, while FL addresses privacy, the challenge of interpretability remains crucial in medical diagnostics. To address these issues, this research introduces FedMed-XAI, a collaborative and trustworthy framework that integrates FL and explainable AI (XAI) within a high-performance computing (HPC)-enabled environment to develop a privacy-preserving and interpretable AI system for skin cancer classification. Using the Skin Cancer MNIST: HAM10000 dataset, our framework was deployed across 20 decentralized clients over 200 communication rounds. The large-scale training and distributed nature of FL necessitate supercomputing resources to efficiently handle parallel processing, real-time performance requirements, and communication overhead across clients. Utilizing distributed HPC power, our framework not only ensures privacy and interpretability but also achieves scalability for real-world medical applications.</p>