Recently, AI-driven discovery in biomedical research is rapidly evolving; an area benefiting from this approach is the creation of a dynamic knowledge system from the integration, processing, organization, and management of biomedical knowledge extracted from text datasets and curated knowledgebases. This thesis presents workflows toward integrating these resources to discover new knowledge in cardiovascular medicine using automated text mining, structured knowledge graphs, explainable biomedical predictions, and generative modeling. The framework cohesively links biomedical entities, identifies hidden relationships, synthesizes new knowledge, and provides actionable insights. Key contributions include uncovering novel protein-disease associations, developing a scalable knowledge graph for multi-modal machine learning applications, and grounding large language model outputs through retrieval-augmented generation. A case study on cardiovascular diseases highlights the applications of this framework to evaluate therapeutics, reveal unexplored connections, generate novel hypotheses, advancing both research and clinical applications. Furthermore, this approach focus on validation and evidence-supported predictions, explores strategies for AI education in biomedical research, bridging technical expertise and domain specific knowledge for effective and responsible AI adoption in biomedicine.