The new era of modern biology and medicine calls for the development of novel integrated approaches leveraging massive multidisciplinary , multi-scale, and multi-modal heterogeneous biomedical data. These biomedical datasets (ex: clinical, imaging, or "omic") are currently stored in data silos [1], which makes their cross-exploitation challenging. Here, we present INEX-MED, a unified Knowledge-Graph based framework that aims to enable systematic integration of diverse data modalities and clinical observations to accelerate both the sta-tistical/semantic data exploitation of biomedical big data with diverse clinical purposes. Following the "FAIR" data principles (Findability, Accessibility, Interoperability, Reusability) [2], we propose a prototype system that (i) integrates clinical, imaging and genomics data from cohorts into a dedicated knowledge-graph, (ii) allows secure access and integrated query processing on heterogeneous biomedical data, and (iii) performs statistical analysis and machine learning to improve the diagnosis/prognosis of studied diseases. The INEX-MED prototype is currently being developed for two use-case cohorts: (i) The ICAN cohort [3] covering 3000 individuals affected by intracranial aneurysm for which clinical records, MRI imaging and exome sequencing data are acquired. The aim is to identify biomarkers and risk factors characterising the development of this disorder. (ii) The MYO-lico cohort with 1200 congenital myopathy patients having clinical records, histopatho-logical imaging data, and exome sequencing data. The aim is to identify novel genes causing congenital myopathies and classify them [4]. INEX-MED already benefits from the Cloud infrastructure provided by the French Bioinformatics Institute (IFB). As a result of this project, the developed prototype will be made available to the biomedical community as an IFB resource, providing methods and technological guidelines to address other biomedical use cases.