Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseInsight · On RadarTNW→ original

Dyania's AI on Nvidia Servers Reads Records of 7,800 Cleveland Clinic Patients

Medical research data is buried in doctors' notes, and teams have spent years reviewing it by hand. Dyania Health's Synapsis AI, running on Nvidia servers inside Cleveland Clinic, read the charts of 7,800 patients who had a replacement heart valve and answered 88 questions about each one. The answers were used to build a model that predicts prosthesis failure at 3, 5, and 10 years. For closed industries, this is a workable pattern for local AI, but the benefit to patients still needs to be proven.

AI-processed from TNW; edited by Hamidun News
Dyania's AI on Nvidia Servers Reads Records of 7,800 Cleveland Clinic Patients
Source: TNW. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTNW ↗
Editor-in-Chief
Key metric #1The records span up to 16 years, with surgeries performed between 2005 and 2021
Key metric #2Every AI answer links to its source record; doctors make the decisions
Key metric #3Cleveland Clinic is an investor in Dyania, so independent verification is needed

The most valuable information in medical charts isn't in neatly structured database fields. It's in free text: doctors' notes, imaging descriptions, discharge summaries. To turn this into research data, clinics have to put specialists to work reading it manually, and that takes years. Now there's an example of AI doing this work. The New Jersey startup Dyania Health has launched its Synapsis AI system on Nvidia servers, inside the American Cleveland Clinic. The system read the charts of more than 7,800 patients who had their aortic valve replaced (the valve at the heart's outflow into the main artery) and answered 88 questions about each one. The data never left the hospital.

What Exactly Was Done

Dyania and Nvidia described the project with Cleveland Clinic. Synapsis AI read the records of patients whose valve replacement took place between 2005 and 2021. For some, the history stretched back 16 years. For each patient, the system answered 88 clinical questions and showed, for each answer, which record it came from.

From these answers, the team built a dataset and a model that estimates the risk of prosthesis failure three, five, and ten years after surgery. According to Nvidia, a manual review of this kind would have taken clinicians years.

The simplest way to picture this is as a very fast research assistant. A doctor doesn't need to open thousands of folders to find out whether a patient had certain symptoms or complications. They ask a question, get an answer, and can check which line in the chart the system relied on. This link to the source matters: without it, the result cannot be audited.

Why Servers Inside the Hospital Matter

Synapsis runs on Nvidia hardware inside the clinic's network, so patient records never leave the building. For healthcare, this isn't a technical detail but a condition for deployment. Medical data is tightly regulated, and most legal and security teams would not allow millions of pages of charts to be sent to an external cloud API.

The practical takeaway for anyone building enterprise AI products in closed industries: deploying on your own hardware is becoming a normal option, not an exotic one. Banks, insurers, and law firms face similar constraints. The cost here isn't tokens paid to a provider, but your own GPUs, their maintenance, and a team to support them. In return, you get control over your data and the ability to pass compliance reviews.

What Has Not Been Proven

It's important not to overstate the result. The study shows that the tool can build a dataset from unstructured records and use it to predict risk. It does not show that patients are better off. Dyania itself acknowledges that it needs external validation. Doctors still check the evidence the AI produces and make all the decisions.

There is also a question of independence. Cleveland Clinic is among Dyania's investors. The clinic had already used the tool last year to select participants for clinical trials. This doesn't invalidate the results, but independent confirmation is especially desirable here.

Meanwhile, the market is accelerating. Google Cloud launched its own product for working with medical records this week, and Bunkerhill raised $55 million. Major players are confidently moving into hospital data. There are still fewer proven improvements for patients than there are announcements.

For developers and product managers, the lesson is this. The value in unstructured data is enormous, but the winner will be the one who solves three problems at once: verifiable answers with links to sources, operation inside the customer's perimeter, and an honest assessment of effect on real outcomes, not just a compelling demo.

Story

Key Questions Answered

What Exactly Was Done?

Dyania and Nvidia described the project with Cleveland Clinic. Synapsis AI read the records of patients whose valve replacement took place between 2005 and 2021. For some, the history stretched back 16 years. For each patient, the system answered 88 clinical questions and showed, for each answer, which record it came from. From these answers, the team built a dataset and a model that estimates the risk of prosthesis failure three, five, and ten years after surgery.

Why Servers Inside the Hospital Matter?

Synapsis runs on Nvidia hardware inside the clinic's network, so patient records never leave the building. For healthcare, this isn't a technical detail but a condition for deployment. Medical data is tightly regulated, and most legal and security teams would not allow millions of pages of charts to be sent to an external cloud API. The practical takeaway for anyone building enterprise AI products in closed industries: deploying on your own hardware is becoming a normal option, not an exotic one.

What Has Not Been Proven?

It's important not to overstate the result. The study shows that the tool can build a dataset from unstructured records and use it to predict risk. It does not show that patients are better off. Dyania itself acknowledges that it needs external validation. Doctors still check the evidence the AI produces and make all the decisions. There is also a question of independence. Cleveland Clinic is among Dyania's investors. The clinic had already used the tool last year to select participants for clinical trials.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…