Fit the context
We design around your team, workflows, and incentives. Solutions are shaped to work in practice from day one.
How we work
Our principles
Right-sized solutions
The best approach is the one that actually gets used. We don't pick the most advanced option for the sake of it.
Uncertainty is a feature
We make uncertainty visible. Honest ranges and clear assumptions lead to better decisions than false precision.
Built to operate, not to demo
Everything we deliver is designed for the people who'll run it on Monday morning, not just the stakeholders who see it on Friday.
Make us unnecessary
If you need us forever, we've done something wrong. We focus on transfer, documentation, and simplicity so your team can own the work.
What to expect
How engaging with us works
Scope
Understand the decision, the constraints, and what good looks like — before writing a line of code.
Build
Develop the right-sized solution: models, tools, or frameworks shaped to your data and operating reality.
Validate
Test against real conditions, stress assumptions, and confirm it works for the people who'll use it.
Transfer
Hand over clean documentation, trained teams, and systems your organisation can run without us.
Where we add value
Problems we solve
Forecasting with sparse or messy data
When historical records are incomplete, inconsistent, or short, but decisions still need a forward view.
Automating decisions under regulatory constraint
Replacing manual processes without losing the audit trail, explainability, or human oversight that compliance demands.
Quantifying downside risk before committing
Understanding what could go wrong — and how badly — before resources, reputation, or safety are on the line.
Prioritising interventions that actually move outcomes
Separating the levers that drive results from the noise that merely correlates, so effort lands where it counts.
Making models useful beyond the demo
Turning proof-of-concept analytics into tools that survive contact with real users, real data drift, and Monday mornings.
Replacing intuition with evidence in high-stakes calls
Giving decision-makers defensible, quantified input when the cost of being wrong is too high to rely on gut feel.
Selected work
Featured projects
Natural Language Processing
Learning from coroners' reports at scale
Extracting structured insight from thousands of unstructured coroners' reports to accelerate public health learning and surface preventable patterns of death.
Read more
Machine Learning
Improving UK-CAT accuracy with machine learning
Improving the accuracy of UK-CAT — a national charity classification system — by training multi-label ML models that tripled classification accuracy over the keyword-matching baseline.
Read more
Clinical Decision Support
Smarter SSRI prescription management
An interactive tool that maps how different antidepressants bind to neuroreceptors, helping clinicians make evidence-informed switching decisions rather than cycling through drugs by trial and error.
Read moreLet's work together.
Tell us what you're working on. No pitch, no commitment — just a conversation.