The team behind Vitrubo.
We build the layer that turns scattered medical documents into one health record people can actually understand, and clinicians can actually use.
Three commitments, held in the architecture.
Glass box transparency
Every interpretation can be inspected: what was stated, and the clinical guideline it rests on.
Numbers are never guessed
Values, units, reference ranges, and abnormality flags are computed by deterministic code. AI reasons only over data that has already been validated.
Identity stays separate from clinical data
The models that reason over results never receive the person's identity.
Product, engineering, clinical, regulatory.
Vitrubo is built by a team spanning product, engineering, clinical, and regulatory backgrounds. Clinical language is reviewed by medical professionals before it ships. Regulatory posture is reviewed by US counsel. We treat both as part of the product, not paperwork around it.
A public benchmark, scored in full.
Stanford MedAgentBench
Public benchmark for medical AI agents.
The team's work in medical AI includes a 100% score on both versions of Stanford's MedAgentBench: 600 of 600 tasks, five runs, zero failures, the first and only perfect score recorded. An active clinical program with Assuta Medical Center, Israel, runs across approximately 100,000 anonymised records, on premises.
The run logs and the technical documentation are available on request.
Request the evidenceTell us about your product.
Integrate Vitrubo into your app, bundle it with your plan, or resell it. Tell us about your product and we will find the shape that fits your ecosystem.