Generative AI is fundamentally reshaping the understanding of biology, medicine, and therapeutics, elevating AI from an analytical tool to a true partner in discovery. Kellis presents work on building foundational AI models that span chemistry, protein structure, gene function, patient states, and therapeutic interventions, moving toward an integrated agentic-AI platform for precision medicine. In the domain of protein structure and function, he introduces ProCyon, a multimodal foundation model that unifies protein sequences, molecular functions, disease associations, therapeutic mechanisms, and structural information. This model enables zero-shot phenotype annotation, drug-binding prediction, and functional interpretation of disease variants, opening the door to annotating the dark proteome and guiding therapeutic targeting. In the field of chemical space modeling, Kellis describes the embedding of molecular structures integrated with global patent databases, drug-target interaction knowledge, and protein function, resulting in a chemically and functionally interpretable drug landscape. This framework supports the generation of novel molecules, functional annotation of the chemical space using co-embedded drug patents, and the discovery of structure-function relationships to advance drug development. For patient trajectory modeling, he presents a latent embedding of patient states derived from multimodal data, including omics, clinical records, imaging, and treatment histories, that enables improved diagnostics, detection of misdiagnosed cases through cross-modal consistency, and the design of personalized interventions tailored to individual trajectories relative to previously treated populations. In the area of personalized therapeutics, Kellis highlights work connecting molecular, cellular, and patient-level phenotypes to predict therapeutic responses and uncover mechanistic disease subtypes. This is achieved through foundational models that integrate molecular function, single-cell expression, protein structure-function data, and chemical information, resulting in AI-driven tools capable of generating causally grounded therapeutic recommendations at the patient-specific level. Finally, he presents Mantis, an AI-powered, human-directed visual data science workbench that allows researchers to explore and interrogate latent embedding landscapes across proteins, chemicals, patients, and therapies. Mantis provides interactive and interpretable visualizations and agentic workflows, where human intuition guides AI actions, revealing latent patterns and advancing the next generation of generative models for biomedicine. Together, these efforts are driving a new paradigm in AI-enabled science, uniting mechanistic interpretability, predictive power, and human-centered discovery to transform biomedical research and precision therapeutics.