In early 2019, two engineers left their respective roles at a logistics firm and a fintech startup because they kept running into the same frustration: companies were buying AI tools they did not need, while the problems that actually called for machine learning went unsolved. So they pooled their savings, rented a desk at a co-working space on Marsh Lane and took on their first client, a Birmingham-based food wholesaler drowning in spreadsheet forecasts that were wrong more often than they were right.
That first project took four weeks. The model we built replaced a manual forecasting process that involved three full-time staff and still missed demand spikes. Within a quarter the wholesaler had cut waste by 18% and reassigned those three people to work that actually required human judgement.
Word spread. A second client followed, then a third. By the end of 2020 we had a team of five and a proper office. Today we are twelve people, still in Birmingham, still asking the same question before every engagement: is AI the right tool here?
These are not slogans on a poster. They are the filters we use when deciding whether to take a project, how to scope it and when to push back on a client request.
If a rule-based system will solve your problem, we will tell you. We have talked clients out of AI projects on the discovery call and pointed them to simpler tools instead. That costs us short-term revenue but earns long-term trust, and those clients come back when they genuinely do need a model.
We do not build black boxes. Every model we deliver comes with documented code, a retraining pipeline and a plain-English explanation of how it works. Your in-house engineers should be able to maintain and improve the system after we leave. If they cannot, we have failed.
A typical project team is two or three people. We work in two-week sprints with a demo at the end of each one. You see progress constantly and can change direction before we have burned through the budget. No six-month disappearing acts followed by a big reveal.
Accuracy on a test set is interesting. Revenue impact is what counts. We define success metrics with you before writing a single line of code, and we track those metrics in production for at least three months after deployment. If the numbers do not move, we fix it on our time.
Twelve people. No middle management. Everyone writes code, talks to clients and reviews each other's pull requests.
Co-founder and ML lead
Former senior data scientist at a Birmingham logistics company. Ravi designs the model architectures and owns the training infrastructure. He has a particular interest in time-series forecasting and spends his weekends restoring a 1970s narrowboat on the canal.
Co-founder and engineering lead
Laura came from a fintech startup where she built real-time fraud detection pipelines. She handles deployment, monitoring and the infrastructure that keeps models running reliably at scale. When a production alert fires at 2 am, Laura is usually the first to respond.
NLP specialist
Daniel joined in 2021 after completing his PhD in computational linguistics at the University of Sheffield. He builds the language models behind our document classification and email triage systems. His thesis focused on low-resource language modelling for West African languages.
Computer vision engineer
Mei spent eight years in automotive manufacturing before joining us. She builds the vision systems that inspect products on factory lines, detecting defects that human inspectors miss at speed. Her models run on edge devices directly on the production floor, avoiding the latency of cloud inference.
Ravi and Laura sign their first client. The demand-forecasting model goes live within five weeks and immediately outperforms the manual process it replaced.
Team grows to five. We move into our own office on Marsh Lane and take on our first computer-vision project for a Midlands-based manufacturer.
Daniel ships our first NLP pipeline, automating document classification for a legal services firm processing 4,000 contracts per month.
We pass 70 completed projects and expand to twelve team members. Client retention hits 97% for the trailing twelve months.
Current focus: helping mid-market companies integrate large language models into internal workflows safely, with proper data governance and cost controls.
Describe the problem. We will tell you honestly whether AI is the right tool, and if it is, what it will take to solve it.
Phone: +44 121 506 7312
Email: [email protected]
Send us a message