Verily researchers trained a novel genomic + EHR multimodal foundation model 10x faster than previous approaches using NVIDIA NeMo Automodel and H100 GPUs

[bernardwin]

Verily and NVIDIA collaborated to add NVIDIA’s GPU-accelerated libraries and frameworks including NeMo Automodel into its Pre platform Workbench which supports nearly 20,000 registered researchers globally.

Press Release: https://verily.com/perspectives/verily-nvidia-precision-health-ai-collaboration

Tech Blog: https://verily.com/perspectives/verily-nvidia-ai-workbench

“As a demonstration of the power of the Pre platform enhanced with NVIDIA AI tools, Verily's researchers have developed the first multimodal foundation model using the NIH All of Us Research Program dataset. The model integrates electronic health records (EHR) and genomics through polygenic risk scores, moving beyond traditional EHR-only approaches to build a more holistic individual health profile, enabling new insights into disease prediction and proactive health management. Using NVIDIA NeMo Automodel and H100 GPUs Verily researchers were able to train the model 10x faster than previous approaches.”