Preemie Collective is the world’s first continuously learning neonatal intelligence platform. It utilises a proprietary federated learning architecture, allowing Neonatal Intensive Care Units to keep patient data private, whilst sharing only anonymised insights.

This platform delivers value to two complementary end-user groups. For Neonatal Intensive Care Units (NICUs) clinical teams, the platform serves as an advanced decision-support system, using early-warning risk models to predict major complications such as BPD, NEC, GF, and length of stay (LoS) for better resource planning. For pharmaceutical and nutrition companies, the platform unlocks historical datasets for multi-centre observational studies, post-market therapy validation, and robust control arms for virtual trials, all without raw data leaving the hospital.

The Preemie Collective is currently under development and is expected to be commercially available by early 2027, promising a transformative approach to neonatal care.

Preemie Collective software features

• The Preemie Collective software is installed within each hospital’s secure IT environment;
• It analyses local, de-identified patient data from the EMR (e.g., Epic, Cerner) to retrain AI models locally;
• Instead of sharing raw data, each hospital’s system sends only encrypted mathematical updates (model weights) to the central server;
• The central server receives encrypted model updates from all participating hospitals in the network and aggregates these updates to retrain the Global AI model;
• This model has learned from the diverse experiences of a vast population of infants, making it more accurate and robust than any model trained on a single hospital’s data.