Dyania Health Deploys On-Premises AI to Analyze Decades of Hospital Patient Records
Powered by NVIDIA hardware and software libraries, Dyania Health's Synapsis AI reasoning engine helps clinical teams extract critical insights from years of unstructured medical records directly behind hospital firewalls.
Hospitals sit on decades of unstructured medical records, ranging from handwritten physicians' notes and operative reports to pathology readings and discharge summaries. Reviewing years of charts manually consumes vast amounts of clinicians' time. To address this bottleneck, New Jersey-based clinical AI startup Dyania Health has introduced Synapsis AI, an assistive clinical reasoning engine designed to analyze complex medical histories without compromising patient privacy.
Clinical Reasoning Directly Behind Hospital Firewalls
Synapsis AI operates across notes, reports, and laboratory results within a patient's electronic health record, connecting disparate findings spanning years. The platform generates traceable clinical signals back to their exact source in the record, alerting care teams to patients requiring urgent intervention.
To meet stringent data security requirements, Synapsis AI runs entirely on-premises using NVIDIA-accelerated computing within a health system's existing firewall. Because all records are processed locally on hospital GPU servers, patient data is never transmitted externally or shared across third-party networks.
Dyania Health develops and trains its clinical models using NVIDIA NeMo open libraries. The models are subsequently optimized for inference with NVIDIA TensorRT and TensorRT-LLM, enabling faster and more resource-efficient execution on local hardware. The system functions strictly as an assistive tool, meaning human clinicians retain final oversight, reviewing the extracted evidence and making all care decisions.
Real-World Validation at Cleveland Clinic
The practical impact of the technology was demonstrated in a major cardiology research initiative at Cleveland Clinic. The medical institution utilized Synapsis AI to support a study examining more than 7,800 aortic valve replacement patients between 2005 and 2021.
Synapsis AI reviewed the complete multi-year cohort—including historical echocardiograms recorded after surgery—a task that would have taken clinicians years to execute by hand. Using this data, the platform helped construct a predictive model assessing heart valve deterioration and the likelihood of reintervention at the three-, five-, and ten-year marks post-surgery.
"Using technology to bring together and analyze information that can help us predict the duration of implanted heart valves is a promising step towards enhanced outcomes and more personalized care for patients," noted Dr. Lars Svensson, chief of the Heart, Vascular & Thoracic Institute at Cleveland Clinic.
Beyond cardiovascular care, the same chart-review framework is being adapted for other complex medical workflows, including matching patients to clinical trials, populating clinical registries, and identifying gaps in ongoing cardiology and oncology treatment.
