Verily Health, working with NVIDIA and the NIH All of Us Research Program, developed an early multimodal foundation model that integrates electronic health records and genomic data—an effort intended to expand beyond single-modality risk prediction. The model is described as learning broad patterns from large datasets before being adapted to specific prediction tasks. Verily’s team argues that EHR data captures one “chapter” of a patient’s story, while genomic information adds inherited disease risk context that may surface before symptoms. The coverage highlights both opportunity and challenge: multimodal integration must overcome differences across data types and resolve how “nature” and “nurture” factors influence each other. For biotech leaders, the effort is an explicit move toward multimodal AI infrastructure for precision medicine, with All of Us providing the scale and diversity typically needed for foundation-model-style training.