SYNAPSE-EU turns scarce, sensitive, siloed data into unlimited training-grade synthetic data — generated on-premise, anonymised by design, and certified for fidelity and privacy before a single sample leaves the vault. The engine behind it is JANUS — one product family, specialised across eleven domains.
A domain-agnostic generative core, delivered as separate, auditable modules.
From raw records to training-grade synthetic datasets at scale — hundreds of thousands to tens of millions of samples — for AI training and model evaluation.
Sensitive information is found and removed whatever shape it arrives in and whichever way it reaches us — so the originals never have to travel or persist.
Full 3D medical volumes — CT and MRI — synthesised on-premise with matching annotations. Extensible beyond medicine.
Raw data enters; only certified synthetic data leaves — each step auditable.
Sensitive originals are stripped and validated on-premise before any modelling.
The generative core produces new samples that match the real distribution.
Utility, fidelity and privacy are measured. Fail the gate, and the batch is rejected.
Each release is sealed with a post-quantum digital signature, so its origin stays provable.
SYNAPSE-EU is the platform. JANUS is the family of products that delivers it — one generative engine specialised per domain, not eleven systems built separately. Four of them have code and tests today; seven are product specifications.
In development — code and tests
Synthetic 3D scans with the annotations models need — and the rare cases reality never supplies in useful numbers. You can ask for the lesion you need.
The trust layer. Records what was generated, from what source, under which guarantees — and seals each release with verifiable proof of provenance.
Finds sensitive information wherever it flows and masks or drops it before it leaves the organisation — on upload, through a connector or API, or live from a database.
The point where policy is actually applied. Without one, a policy is a document. Hardened against deliberate attack, not merely tested for correctness.
In specification — no code yet
Every release is measured against four independent bars before it is trusted.
Models trained purely on synthetic data are tested against real data; we target near-parity with training on the originals.
Beyond looking right: distributions, correlations and structure are checked to match the source.
Rigorous privacy checks ensure synthetic samples never reproduce real individuals. Privacy is proven, not assumed.
A post-quantum signed Data Passport records how each dataset was made, so authenticity can be verified independently — by anyone, at any time.
The things teams ask us first, answered plainly.
SYNAPSE-EU is a sovereign synthetic data platform by CAI Technology. It turns scarce, sensitive or siloed data into unlimited training-grade synthetic data, generated on-premise, anonymised by design, and certified for fidelity and privacy before any sample is released.
Synthetic data generation across tabular, signal, image and text modalities; data anonymisation with de-identification built in and re-identification risk measured rather than assumed; and imaging synthesis producing full 3D medical volumes such as CT and MRI with matching annotations.
No. SYNAPSE-EU runs fully on-premise on European infrastructure, with no third-party clouds and no data egress. Sensitive originals are de-identified and validated on-premise before any modelling takes place.
Every release passes a quality gate measuring four independent bars: utility (models trained on synthetic data are tested against real data, targeting near-parity), fidelity (distributions, correlations and structure are checked against the source), privacy (rigorous checks ensure no real individual is reproduced), and provenance. A batch that fails the gate is rejected.
Each dataset release is sealed with a Data Passport recording how it was made, signed with a hybrid signature: a NIST-standardised post-quantum algorithm combined with a proven classical one. Verification succeeds only if both hold, so provenance stays verifiable both today and after the quantum transition.
Medical imaging is the domain taken furthest, and the three transversal products — the trust layer, anonymisation in motion and the enforcement point — apply to any data. Geospatial, energy, IoT, industrial control, counter-UAS, RF and text are product specifications built on the same engine, not deployable products yet.
SYNAPSE-EU is the platform; JANUS is the family of products that delivers it. One generative engine, specialised per domain, rather than eleven systems built separately. What changes between domains is the kind of data and what you condition on — the engine, the validation discipline and the guarantees stay the same.
Four have code and tests: Med, the medical imaging vertical, which is the most advanced; Core, the trust layer; Data, which anonymises information in motion; and Edge, the enforcement point. The other seven — Geo, Energy, IoT, OT, UAS, RF and Text — are product specifications. We do not present them as deliverables.
SYNAPSE-EU runs fully on-premise — European infrastructure, no third-party clouds, no data egress. Built for regulated, high-stakes domains.
Pilots, partnerships, data collaboration — technical questions welcome. We reply within 24 business hours.