Models that understand
the underlying language of biology

We build Large Medicine Models that represent human biology with unprecedented fidelity, drawing on large-scale, rich, multimodal data to identify complex processes opaque to conventional analysis.

Our models operate in the real world and learn directly from it. They are shaped by data from the clinical stream, and are designed to handle its noise, incompleteness, and heterogeneity.

We model signals at the finest available scale with novel AI architectures of matched expressivity, capturing the complex, widely distributed interactions that define the biological realm.

Our focus is representational: revealing the fundamental biological structure that lies behind the data as the critical foundation for all downstream objectives in healthcare.

Real clinical data

Hologen trains on decades of real clinical data from millions of real patients: imaging, blood tests, genetics, clinical notes and outcomes.

These records hold the detail of how disease behaves and how people respond. Learning from that detail is the starting point for individual predictions.

Multimodal foundation models

We build one foundation model per modality, covering its tasks, diseases and body parts. Further models combine these modalities to represent the patient as a whole.

Each task is evaluated against specialist systems, from image reconstruction and harmonisation to biomarker estimation, diagnosis and prognosis.

Held to medical device standards

All of this work is done in a manner compliant with ISO 13485, the international standard for quality management of medical devices. Clinical and scientific teams can build on models that are versioned, reproducible and documented.

From insight to intervention

The models sharpen diagnosis in the clinic, strengthen trial design, select the patients most likely to respond, and help prevent disease through earlier intervention.