About our Artificial Intelligence practice
We started in 2021 with a single client and a rented GPU. Four years later we are a team of nine, and our models process over two million data points every day for businesses across the UK.
How we got here
Our founder, Marcus Hale, spent eight years building internal machine-learning tools at a large UK insurer. The models worked well. Getting them approved, documented, and into production was the hard part. Most of the value was lost in translation between the data team and the rest of the business.
He left in early 2021 to start a consultancy that would treat deployment and explainability as first-class deliverables, not afterthoughts. The first project was a fraud-detection system for a regional building society. It went live in six weeks and reduced false-positive alerts by 34%.
Word spread. By the end of 2022 we had hired three more engineers and moved into a permanent office in Long Carroll-Hane Park. Today we work with logistics firms, SaaS platforms, healthcare providers, and energy companies. The common thread is that each client has data they know is valuable but have not yet been able to act on.
We remain self-funded. No venture capital, no pressure to chase revenue targets that conflict with doing good work.
What guides our decisions
These are not slogans on a poster. They are the criteria we use when scoping projects, hiring people, and choosing which clients to work with.
Honest scoping
If a spreadsheet formula or a simple rule engine would solve the problem, we say so. We have turned down projects where AI was not the right tool, and we have recommended competitors when their specialism was a better fit. Trust matters more than a single invoice.
Transparent pricing
Every project gets a fixed quote before work begins. If we underestimate the effort, that is our problem, not yours. We publish our rate bands on the pricing page so there are no surprises in the first call.
Explainability by default
Every model we ship comes with documentation that a non-technical stakeholder can read. We include feature-importance charts, example predictions with reasoning, and a limitations section that describes when the model is likely to be wrong.
Measured outcomes
Before we write code, we agree on a success metric with you: accuracy percentage, time saved per week, cost reduction in pounds. At handover we measure it together. If the metric is not met, we keep iterating at no extra charge until it is or we refund the difference.
The people behind the models
Nine full-time staff, no contractors. Everyone listed here has been with us for at least a year.
Marcus Hale
Founder and lead engineer. Eight years at Aviva before starting the company. Specialises in time-series forecasting.
Anisha Kapoor
NLP lead. PhD in computational linguistics from Edinburgh. Built our document-processing pipeline from scratch.
David Osei
Infrastructure lead. Manages our UK-based GPU cluster and CI/CD pipelines. Previously at AWS for six years.
Sophie Brennan
Project manager. Keeps timelines honest and clients informed. Certified Scrum Master with a background in fintech delivery.
Key milestones
A short history of the company so far.
March 2021
Company registered. First project signed with a regional building society. Fraud-detection model goes live within six weeks.
November 2021
Second and third clients onboarded: a parcel-logistics firm and an online retailer. Team grows to four people.
June 2022
Moved into permanent office space at 2 Shields Terrace, Long Carroll-Hane Park. Installed our own GPU training cluster.
January 2023
Completed our first healthcare project: a diagnostic-triage model for a private clinic chain. Passed external audit on first submission.
September 2024
Reached 47 production models across 28 clients. Team size: nine full-time staff. Annual retraining programme launched for all active models.