About us

Every model we build is trained on a client's own data, for a specific decision that client needs to make.

The work starts before any modelling: understanding the decision, the data that's actually available, and what a good outcome looks like in practice. Sometimes the most useful thing we deliver early is a clear read on whether machine learning is the right tool for the problem at all.

Our work runs across structured prediction, computer vision, forecasting, experimentation, dataset work, and reinforcement learning. That breadth means we can draw on what has worked on adjacent problems.

We work alongside your team rather than at arm's length: sharing assumptions, showing intermediate results, and handing over models, code, and documentation your people can maintain once the engagement ends. Or we can keep an eye on things if that's what you'd prefer.

Client work is kept confidential. This site describes the kinds of problems we take on rather than naming who we've done them for. We don't resell general-purpose AI or off-the-shelf LLM products. We apply industry methods and architectures to help find your models and solutions.