What You Should Know:
– Verily, a precision health AI company, today announced a collaboration with NVIDIA to integrate NVIDIA’s complete AI technology stack into its Pre platform.
– The strategic integration is designed to help Verily’s health system, life science, payer, and government customers and partners more efficiently develop and deploy AI across healthcare.
Unparalleled Acceleration for Omics and AI Development
The collaboration provides researchers with immediate access to cutting-edge tools and hardware within Verily’s trusted research environment (TRE), Workbench.
- GPU Integration: Verily has added NVIDIA’s GPU-accelerated libraries and frameworks, including NVIDIA NeMo, NVIDIA Parabricks, and NVIDIA CUDA-X Data Science. Researchers now have access to some of the latest GPUs, including NVIDIA Blackwell and NVIDIA Hopper.
- Speed and Efficiency: This integration has demonstrated the ability to greatly accelerate analyses, from hours to minutes, when using Parabricks and B200 to process genomic data compared to traditional, CPU-only methods.
“With NVIDIA’s cutting-edge AI capabilities now available in our Pre platform, we are providing researchers with unparalleled tools to enhance the speed and efficiency of AI model development and omics analysis,” said Stephen Gillett, Verily CEO.
Powering the NIH All of Us Research Program
A critical element of this effort is the acceleration of analyses within the National Institutes of Health’s (NIH) All of Us Researcher Workbench. This Workbench, which hosts one of the world’s largest genomics datasets and supports nearly 20,000 registered researchers globally, which will be powered by the Pre platform through Verily’s partnership with Vanderbilt University Medical Center.
As a demonstration of the platform’s enhanced capability, Verily’s researchers developed the first multimodal foundation model using the NIH All of Us Research Program dataset. This model integrates electronic health records (EHR) and genomics through polygenic risk scores, enabling new insights into disease prediction and proactive health management. Researchers were able to train the model 10x faster than previous approaches using NVIDIA NeMo Automodel and H100 GPUs.
Expanding AI Capabilities Across the Pre Platform
Beyond the Workbench integration, Verily plans to make NVIDIA GPU-accelerated tools available across other Pre solutions, enhancing data management and exchange.
- Refinery: This powerful curation engine ingests, harmonizes, and enriches multimodal data, using a clinically informed, FHIR-native data model to transform unstructured data into structured data assets.
- Exchange: A secure forum for sharing and accessing AI-ready datasets, models, and agents to advance precision research and care.
