Pharmacogenomics are increasingly used in precision medicine for the development of targeted drug therapy to personalize treatment to enhance treatment outcomes and minimize adverse drug reactions (ADRs). In this study, we introduced an AI-based multi-omics model that integrates and amalgamates different data types including genomics, transcriptomics, epigenomics, and clinical data as predictors of patient-specific score of drug response. The included PharmGKB source of drug response offered the opportunity to use GNNs, advanced multi-modal data fusion, and dimensionality reduction to model complex biological interactions. Predictive supervised learning models were trained on a high number of features with high predictive accuracy, while reinforcement learning and digital twin through in silico validation were mapped to simpler personalized dose optimization. The hybrid model showed excellent capacity with 97.2% prediction accuracy, AUC 0.97, and full interpretability (the best-performing model of existing high-performing models at the time of submission). These empirical results provide evidence of the framework’s capacity to interrogate key pharmacogenetic features and effectively stratify patients into groups so as to develop a more effective and safer drug regimen. To summarize, the integrative AI platform supports a framework for the implementation of precision pharmacotherapy, allowing for a higher degree of data-driven decision making and treatment optimization based on personalized parameters, all judged to be credible by the patient.