Two predictive ML models deployed inside the NHS.
Built by a data science team stood up from scratch.

Hospitals don’t move at research speed.

The company had promising machine learning work in clinical prediction, but getting it into a hospital environment is a different problem to building it in a lab. Complex clinical settings, existing systems, and clinical staff all had to be worked around, not just designed for.

On top of that, there was licensable IP coming out of academic research that needed evaluating before anyone could commit to it, and every model built had to clear a clinical and regulatory bar before it could be trusted near patients.

Three constraints were non-negotiable:

• Deploying AI inside complex hospital environments
• Evaluating and integrating licensable IP from academic research
• Meeting the clinical applicability and regulatory bar for ML tools

Two models, one new team, one roadmap.

Two clinically relevant models were built using advanced ML techniques: a COPD exacerbation prediction model and an ICU early warning system, both designed to flag deteriorating patients earlier than existing processes.

Delivering them meant building the team first. A six-person data science team was recruited and led from the ground up, with its own product roadmap rather than being bolted onto an existing one.

In parallel, neural network and Bayesian methods were evaluated for their potential to accelerate drug discovery, feeding into the company’s wider view of where academic IP and partnerships could add the most value.

Two models live. A team built. A pathway to product.

Two clinically relevant models went from research into NHS-linked pilots, backed by a data science team built specifically to deliver them. Beyond the pilots, the work set out how the company evaluates academic IP and partnerships, and the pathway for turning validated models into real products.