We build AI systems with clinical and research teams, from unsolved genetic cases to the data platforms under them. A person signs every decision.
Products for clinics, models for the science, and the platform under both.
Workbenches that assemble a case, rank what to look at next and explain why. Built with the clinicians who use them.
Phenotype, variant and discovery models, plus agents that investigate under supervision. Cited everywhere. Tested on published cases.
Versioned pipelines from raw reads to results, curated knowledge releases, audit trails. Each organisation’s data stays its own.
Our first product, in a limited pilot with clinical genetics teams.
Six models investigate. Every rank cites its record. A clinician decides.
Built with the people who use it, tested on published cases, and a person signs. The same on every project.
A clinician or scientist from the partner team sits in every weekly review. We ship what they would sign.
Held out, knowledge frozen to the case date, limits written down. Validation summary on request.
Our systems propose, explain and record. No gate passes without a person. Nothing we build gives a diagnosis.