Chip Giant Nvidia Partners with Abridge to Develop Healthcare-Specific AI Model
Chip giant Nvidia is collaborating with startup Abridge to create a baseline model specifically designed for clinical conversations. This AI model aims to enhance the accuracy, reliability, auditability, and customization of clinical workflows, covering areas from documentation and evidence preservation to workflow automation and clinical reasoning support, the companies announced recently.
Development of the Clinical AI Model
Building upon the Nvidia Nemotron open model family, where model weights and training data are readily available, the new Abridge model is trained on Nvidia’s Blackwell AI infrastructure using advanced pre-, mid-, and post-training processes with anonymized data, according to the companies.
Abridge leaders highlight that training in all three phases will allow clinical knowledge to be embedded from the ground up, thus improving accuracy, precision, and reliability across various specialties, care settings, and the multi-step workflows that follow a clinical encounter. By adapting the domain earlier in the training lifecycle, Abridge can create a model that is clinically justified at its core. Nvidia’s Nemotron gives Abridge the ability to optimize quality, cost, and efficiency at every level, delivering the right model for the right workflow at the right scale, executives explained.
Abridge’s Expanding AI Platform
Nvidia is also an investor in Abridge through its venture capital arm, NVentures. Abridge is rapidly expanding its AI platform beyond just an AI writing tool to act as a comprehensive AI clinical assistant. This week, the company announced a platform expansion to integrate payer and life science workflows. Described as an “AI-native clinician intelligence platform,” Abridge claims it now connects care delivery, payment, and evidence-based treatment.
The company collaborates with over 300 healthcare systems, supporting more than 100 million conversations annually. As Abridge expands its AI platform, it aims to enhance the performance and speed of its models.
“You want to distribute information everywhere, and how do you figure out how to do that? How do you also figure out how to deliver the right level of accuracy and the right latency? You end up needing a little more control in certain places than you might have expected, and we’ve had the incredible privilege of working together and building on the Nemotron family of models,” said Shiv Rao, CEO and co-founder of Abridge, during a keynote event with Kimberly Powell, vice president of healthcare at Nvidia, in New York City.
Broader Collaborations in Healthcare and Life Sciences
Health technology companies see significant opportunities to leverage Nvidia’s computing power to enhance their AI capabilities. Last fall, Verily, part of Alphabet and Google’s life sciences sister company, announced a collaboration with Nvidia to integrate its AI technology stack into Verily’s pre-platform. Innovaccer also announced that it has acquired Nvidia’s full-stack AI platform to accelerate its voice, text, and multimodal reasoning infrastructure and power its AI agents.
The computer chip giant is also advancing into life sciences. Eli Lilly and Roche have entered an AI infrastructure collaboration with Nvidia, Fierce Biotech reported. In medical technology, the company is working with Thermo Fisher Scientific to expand its autonomous laboratory infrastructure and with Qiagen, a Netherlands-based diagnostics manufacturer, to improve researchers’ ability to use AI in the drug development process, Fierce Medtech reported. Through collaboration with diagnostics manufacturer Droplet Biosciences, Nvidia is delving further into medical technology and cancer research.
Nvidia is optimistic about healthcare, with Powell stating that it “will be one of the largest industries in technology.”
“Nvidia is not a healthcare company; we never will be and we never want to be, but to enable us to tackle some of the toughest jobs in the world, jobs that have enormous impact, we believe we have a unique ability to contribute,” Powell expressed.
The Future of AI in Healthcare
In the last 18 months, the tech industry has experienced “three major technological breakthroughs,” from “AI that can create things, to AI that can reason through things, to AI that can work,” Powell noted. There is a growing need for AI intelligence that is more domain-specific and tailored for complex healthcare workflows, as generic AI models “do not understand clinical language, do not have clinical reasoning, and certainly do not have the domain expertise of all the tedious tasks and interconnected work required for workflows to be fully transformed,” she explained.
“Any technological breakthrough requires looking at the entire stack. We don’t look at AI as a model, but as a comprehensive computational problem. It starts with energy, then it’s chips and systems, then it’s AI data centers and clouds, then it’s the core foundational models, and finally it’s the application layer.” Abridge has made a strong case that the vertical AI application layer has immense value and effectiveness, Powell remarked. “We collectively realized that it’s time to go deeper into the stack – a clinical conversation foundation model so that the complexity of healthcare and all the workflows and the connectivity of this amazing ecosystem that you’ve brought to fruition can be realized, because it needs to become much more domain specific.”
For more information, visit the source.
“`
