Applied research that turns into shipping product.

The Centre runs a small number of multi-year research programmes alongside a continuous flow of shorter focused projects. Current programmes include forecasting at the edge of stationarity — methods for time series whose statistical properties shift more often than classical models assume; lightweight computer-vision models for industrial inspection on commodity hardware; hybrid neuro-symbolic decision systems for regulated domains; multilingual sentiment under domain shift; and the operational economics of generative-AI workflows in enterprise environments.

Each programme has a written multi-year plan, a designated principal researcher, a small core team and a clear funding model. Plans are revisited annually and rewritten when the field moves enough to warrant it. We have closed programmes when the questions they were addressing turned out to be less consequential than they appeared, and we expect to do so again; the willingness to close programmes is what keeps the open ones honest.

Shorter projects — typically three to nine months — fill the space between programmes. They include applied investigations on behalf of specific industrial partners, contributions to open-source projects we depend on, and the occasional pure-curiosity sprint that has not yet earned a programme of its own.

The Centre publishes peer-reviewed papers, technical reports, open-source code and the occasional public dataset. We are deliberate about the venues we publish in, with a preference for venues where the audience is composed of practitioners and the review is rigorous. Publication is not the goal of the work; it is one of several outputs, alongside production deployment, internal capability and partner enablement.

Open-source contributions are a substantial part of our output. We maintain libraries that began as internal tooling and that we now consider too useful — and too costly to maintain alone — to keep proprietary. We contribute upstream to the data, ML and visualisation projects we depend on, and we participate in the governance of a small number of foundations whose work intersects with ours.

Production deployment is the output we hold ourselves to most strictly. A piece of research that does not reach a real user, in a real deployment, with a real measurement of its effect, is treated internally as incomplete rather than as a success. The discipline is unglamorous and is the principal reason the Centre is funded.

“An idea that does not survive contact with production is not yet a result.”

The Centre operates joint laboratories with several universities, with shared researchers, shared funding and shared intellectual-property arrangements. The relationships are long enough — typically five to ten years — to outlast any single project, and they include explicit commitments around doctoral funding, internships and research mobility in both directions.

Industrial partners co-fund roughly half the Centre's annual programme, with arrangements that vary from full sponsorship of a specific research stream to participation in a shared consortium. Partners have early access to results that affect their domain, the right to seat researchers in their teams during deployment, and the option to take licenses on resulting IP under terms agreed in advance. We are deliberate about not allowing any single partner to dominate the Centre's agenda; the breadth of partnership is what keeps the work generally useful.

Beyond the core team and the formal partnerships, the Centre maintains a network of affiliate researchers, practitioners and senior advisors across Europe, the Middle East, Africa and Asia. Affiliates participate in seminars, contribute to specific projects, co-author papers, and provide candid critique on work in progress. The network exists because the questions we want to answer are too varied to be answered well by any single team in any single geography.

Membership of the network is by invitation, but the door is wide. We are particularly interested in mid-career and senior practitioners with deep applied experience, doctoral candidates working on questions adjacent to ours, and industry leaders willing to expose hard real-world problems to research scrutiny. Joining instructions are on the dedicated application page.

The Centre is funded from three sources, in roughly equal proportions across a normal year. The firm contributes from operating revenue; industrial partners contribute through programme sponsorship and consortium membership; and a small share comes from public research funding where the project structure makes that appropriate. We do not accept funding that would constrain our ability to publish negative results, and we do not accept funding from sources whose interest in the outcome is incompatible with the integrity of the work.

Detailed accounts of the Centre's funding and outputs are published annually for partner organisations, and on request for any party with a legitimate interest. Transparency on funding is, in our view, a precondition for credibility on the work itself.