The Pulse

Dyania and Nvidia team up on clinical AI inside hospital networks

Dyania Health and Nvidia will work to bring Synapsis AI to health systems using computing infrastructure inside their networks. A Cleveland Clinic study used the software to analyze records of 7,812 aortic valve replacement patients.

AI.info Team · · fact-checked · against 2 sources ↓ · Published 13 Nov 2026

Dyania Health announced a collaboration with Nvidia on Nov. 13 to put clinical AI that analyzes patient records to work on health systems’ own infrastructure. Dyania says Nvidia accelerated computing lets its Synapsis AI process records behind the hospital firewall without transferring them outside the health system.

The software is designed to answer clinical questions using evidence from a patient’s chart, including notes, operative and imaging reports, and laboratory results. The collaboration’s first work is a Cleveland Clinic study involving 7,812 patients who received aortic valve replacements between 2005 and 2021.

Records stay within the network

Synapsis AI is built to examine records over time and connect clinical information across documents. Dyania says much of the information health systems need is in narrative text rather than structured fields, making it difficult to analyze without manual review.

Nvidia describes the product as an assistive tool: it surfaces evidence from medical records that a clinician must confirm and act on. Its output consists of clinical signals traceable to their sources in the record, with clinicians reviewing the evidence and making decisions.

Dyania trains its proprietary models using Nvidia NeMo for model training and reinforcement learning. Within a health system, Synapsis AI runs on Nvidia TensorRT and TensorRT-LLM using the hospital’s own Nvidia accelerated computing infrastructure; inference stays on the private network.

The partners plan to bring the system to health providers in the United States and internationally, publish joint case studies and integrate it with Nvidia’s healthcare software stack. They are also evaluating Nvidia Confidential Computing for training and inference, Nvidia Dynamo for distributed inference serving, and Multi-Instance GPU for dividing GPUs among concurrent clinical workloads.

Another area under evaluation is scaling to many Nvidia Blackwell Ultra GPUs as deployments move from batch processing toward near-real-time clinical insight at the point of care. Those areas are under evaluation, rather than completed parts of the collaboration.

Heart valve study

The Cleveland Clinic study examined how long replacement aortic valves last and whether patients would need another procedure. Researchers reviewed up to 16 years of patient care, much of it documented in narrative notes and echocardiogram reports.

Running on Nvidia computing inside Cleveland Clinic’s environment, Synapsis AI identified the study population and answered 88 clinical questions for each patient. Those questions covered valve characteristics, co-existing conditions, laboratory trends, medications, follow-up ultrasound measurements and later clinical events.

The study produced a clinical dataset in far less time than a manual review would take, according to Dyania. A predictive machine-learning model was also trained to estimate timelines for valve failure and the risk of another intervention.

Dyania says the model found that the type of implanted valve was the strongest determinant of how quickly it deteriorated, while age and other medical conditions affected whether and when a second procedure occurred. The analysis also examined how follow-up heart ultrasounds can improve predictions of future interventions.

The model estimates risk at three, five and 10 years and updates its outlook with each new ultrasound. The tool is used to help inform care but requires external validation from clinicians.