R&D lab
Some work takes months, not days: modernising legacy code, security work, and experiments in robotics, vision models, machine learning and space. Our R&D floor takes it on with you, in the open or behind closed doors.
You choose how visible the work is. The rigour is the same either way.
We run the work as a public experiment.
The plan is written down and frozen before we start. The results are published, including the ones that do not go our way, and the code is released as open source. Anyone can check the work, and your name can be on it.
Private, confidential work.
We work in your repositories and on your infrastructure, and nothing is published unless you decide to. Confidentiality and who owns the results are agreed in the proposal. The method is the same: a written plan, checks that run, and evidence you can inspect.
Four kinds of work, each with a way to check the result.
Old systems that still run the business. We work out what the code actually does, pin that behaviour down with checks, and then refactor in small steps that are each verified against the old behaviour.
See the migration witness demoReviews of code, access and infrastructure, followed by fixes that land with tests. Where it makes sense, the checks go into your pipeline so they keep running on every change.
How we handle controlsRobotics, vision models, machine learning and space: questions where nobody knows the answer yet. We design the experiment, build the prototype and measure what happens, so you decide on results rather than on hope.
Browse our public experimentsYour architecture rules, written down as code. Set up as a CI gate, our open-source Blueprint Conformance Engine fails the check when it detects a breach of a supported rule, so the rule doesn’t depend on a reviewer remembering it.
Read about the engineEngagements run over months. We scope them together and agree each one by proposal, so there is no price list here.
What you want to find out, build or fix, and whether the work should run in the open or behind closed doors.
Scope, plan, what counts as done and how we will measure it. You see it before anything starts.
We work in cycles agreed in the proposal. Each cycle ends with something you can run or check, and a decision about the next one.
Code, evidence and documentation, handed over as agreed in the proposal. In the open, the results are published as well.
Prefer a shorter, fixed-price start? See the AI Innovation Sprint.
