NVIDIA helps AI sift through decades of patient charts so doctors don't have to

Healthcare AI startup Dyania Health has developed Synapsis AI, a clinical reasoning engine designed to process a patient's entire medical history, which is often spread across decades of unstructured text like doctors' notes and lab reports. The system, powered by NVIDIA's NeMo and TensorRT-LLM technologies, runs on hospital servers behind their firewalls to keep patient data secure, and it works by connecting disparate findings across years to surface relevant clinical signals that need a clinician's attention.

For hospital teams, this means an end to manually sifting through years of patient charts for studies or routine care. Clinicians get a summarized, evidence-backed list of patients who need intervention and why, with every signal traceable to its source in the original record. The real-world impact is less administrative burden and more face-to-face patient time, as demonstrated in a Cleveland Clinic study reviewing thousands of aortic valve replacement patients. The takeaway is that specialized AI is becoming a practical, secure tool for turning historical patient data into actionable insights without compromising clinician oversight.