The product family

One engine.
Eleven domains.

SYNAPSE-EU is the platform. JANUS is the family of products that delivers it. Four of them have code and tests today. Seven are product specifications built on the same engine — and we say so on the page rather than in a footnote.

The four with code → What is measured
The common recipe

How data gets synthesised, whatever the domain.

The same three beats whether the data is a tomography, an electricity demand curve or a radio signal. What changes between domains is the kind of data and what we condition on. The engine, the validation discipline and the guarantees stay the same.

01
Compress

Keep the structure, drop the bulk

A model learns to hold real data in a far more compact form while preserving what matters. For a scan that means the anatomy; for an electrical grid, the demand patterns and how they depend on each other. This is what makes generation possible at scale.

02
Generate

Ask for what you actually need

A second model creates new compact representations, refining noise step by step, guided by the request: a lesion on this artery, a grid failure at peak hour, a vibration anomaly on the front bearing. Each result arrives with its label, because both are produced together rather than separately.

03
Verify

Three proofs, or it does not ship

That it resembles reality statistically. That a model trained only on the synthetic data works on real data. And that nobody in the original set can be recognised or reconstructed from what was generated. Fail one, and the batch is rejected.

In development

Four products with code and tests.

One vertical taken deep, and three transversal products that apply to any data.

Advanced

JANUS Med

Train medical AI without touching a patient.

Real medical data is scarce, sensitive and hard to share: a diagnostic algorithm needs thousands of annotated cases, a hospital produces a few hundred a year, and those are protected by GDPR and an ethics committee. JANUS Med generates CT and MRI volumes that look and behave like real ones, together with the segmentations models need, without corresponding to any existing patient.

You can ask for exactly what you need: a lesion of a given size, on a given artery, ruptured or not, in a patient of a given age. That is what solves the rare-case problem — a pathology with forty cases a year will never produce enough real data, but it can be generated in numbers sufficient for training. Every delivered set comes with a validation report: how close it is to reality, whether a model trained on it works on real patients, and proof that no original case can be reconstructed.

For medical software vendors · hospitals · clinical research · regulators
In development

JANUS Core

The trust layer.

A synthetic dataset without evidence is just a large file. JANUS Core is the component that turns each delivery into something a compliance officer can accept on argument rather than on faith. It keeps the record of what was generated, from which source, with which parameters and under which guarantees — and seals each set with verifiable proof of provenance.

It administers who has access to what, keeps the complete log of operations, and issues the certificate that travels with the data when it moves on. For organisations working with regulated data, this is the part that separates "we generated some data" from "I can demonstrate at any audit how it was generated, and that it exposes no one". It runs entirely inside the client's own environment.

For security directors · data protection officers · compliance teams
In development

JANUS Data

Protects data in motion.

Most data leaks do not come from spectacular attacks but from ordinary flows: an export to a supplier, a file sent by email, an integration between two systems. JANUS Data sees where sensitive information travels inside an organisation and masks or removes it at the moment it moves — on bulk upload, through a connector or API, or live from a database — so personal and confidential data does not leave the place it is allowed to be.

Detection is automatic, by type of information, not by rules written by hand for each field. The sensitive originals stay put and what travels onward is already clean. It is also the component that feeds the generation engine: whatever enters training has already passed through anonymisation.

For security and data teams · organisations with compliance obligations
In development

JANUS Edge

The enforcement point.

Data protection policies are worth something only if someone actually applies them, at the point the data passes through. JANUS Edge is that point: the gate information travels through between systems, applying the rules set in Core before anything moves on. Without such a point, a policy stays a document.

It has been through an adversarial security review — not merely tests confirming that it works, but deliberate attempts to break it, with everything found fixed before any real use. It is designed to run inside the organisation, with no external dependencies.

For infrastructure and security teams that must guarantee nothing sensitive leaves unintentionally
Evidence

The numbers we have, and the claims we do not make.

Medical imaging is the first domain taken all the way, so it is the one with measured figures. The rest is stated as what it is.

Measured today

0.82 reconstruction fidelity across 2,600 volumes.

The reference open model (NVIDIA MAISI) reaches 0.84 — carrying four times the parameters and trained on roughly four times the data. The generator itself is in its first full training run at the time of writing.

  • Built but not calibrated. The quality gate exists and works — it correctly rejected a deliberately weak model — but its thresholds are provisional and have to be calibrated on real data before they mean anything in public.
  • What we do not claim. That the generated data has been clinically validated, that a medical device trained on it has passed any certification, or that the proposed verticals have been demonstrated. Seven of the eleven are product specifications with a shared engine proven on the first domain — not shippable products.
  • The real differentiator is not image quality. Models that generate medical data exist. Very few can demonstrate numerically that from what they generated you cannot get back to a real person. For regulated data, that is the argument that opens an ethics committee's door.
Specification

Seven more, on paper.

Product specifications on the same engine. No code yet, no delivery date, and nothing here is for sale. They are listed because the engine is domain-agnostic and these are the domains where the shortage of shareable data is worst.

Specification

JANUS Geo

Models reading satellite imagery or cadastral maps need complete coverage: every season, every terrain, every rare event. Would generate Earth-observation imagery and cadastral records — floods, sudden vegetation change, unauthorised construction — with the classifications already attached.

geospatial · cadastre · agriculture
Specification

JANUS Energy

A grid operator's operational data is among the most sensitive it holds, and exactly what is needed to train forecasting models. Would generate demand, generation and grid behaviour, including the stress scenarios an operator cannot produce on request and would never share.

grids · generation · demand
Specification

JANUS IoT

A device fleet produces enormous volumes of data but almost never the part you need: the anomalies. Would generate realistic telemetry, normal and abnormal, with the anomalies already labelled, so predictive-maintenance models see every case without waiting for it to happen.

sensors · smart buildings · telematics
Specification

JANUS OT

Industrial plant data is sensitive twice over: it reveals process know-how and it is safety-critical. Would generate process and control data including fault conditions, so algorithms can be built and hardened without the plant exposing how it actually runs.

industry · critical infrastructure
Specification

JANUS UAS

Drone detection has to work first time, in a situation that occurs rarely and cannot be repeated on request. Would generate the sensor signatures used for detection and tracking, including combinations reality supplies only occasionally.

defence · site security · airports
Specification

JANUS RF

Capturing real radio scenarios is expensive and, for the cases that matter most, close to impossible. Would create signal and spectrum data covering exactly the events real recordings do not — the ones an operator meets once every few years.

defence · telecom · spectrum monitoring
Specification

JANUS Text

Language models trained on regulated domains have a structural problem: the very texts they need contain information about real people. Would produce documents, reports and records that read like authentic ones, without any real person in them.

regulated documents · health · legal
Sovereign by design

Bring a representative set, and we measure on your data.

The honest way to find out whether this works for your domain is not a brochure. Tell us what you are trying to train and what you are not allowed to share.

Talk to us →