The capabilities below are built and running on the platform today. Not a roadmap — the working system.
Four stages, end to end. Your data never leaves the institution that holds it — only governed, verifiable outputs do.
Access control, disclosure control, and audit are mechanisms the platform runs — not promises in a policy document.
A governed catalogue of reactivated ophthalmic and neuro cohorts — each with provenance, access terms, and institution-scoped isolation, with every request logged.
Federated learning runs the model inside each institution over a signed, outbound-only node. Only model updates leave — never the images. Fine-tune modern foundation models on cohorts that were previously unreachable.
Differentially private queries with an enforced per-dataset budget, synthetic datasets, and automatic small-cell suppression — the numbers you take away can't be traced back to a person.
Every result gets a signed Verifiable Research Certificate binding it to its lineage, disclosure settings, and a tamper-evident audit trail. Verifies offline against a public key — no account, no contacting us.
OMOP CDM v5.4 with the imaging extension, FHIR R4 ingestion, and one-click evidence packs that support your institution's compliance assessments under the UK (Five Safes · DSPT), EU (GDPR · EHDS), and US (HIPAA · Common Rule) frameworks.
Federated analysis isn't a bet. Models trained this way have reached roughly 99% of centralised performance in multi-institutional brain-imaging studies (Sheller et al., Scientific Reports, 2020), and a single clinical model has been trained across 20 institutions worldwide without any data leaving a hospital (Dayan et al., Nature Medicine, 2021).
For a data custodian, participation never creates a new copy of the dataset anywhere — which shortens governance review, keeps the institution in control, and aligns with UK GDPR Article 89 research safeguards and the FAIR principles UK funders now expect.
We're onboarding data custodians and research partners now. Tell us what you're holding or what you're trying to find out.