Join the Research Center

The network has three groups of members in roughly equal proportions. The first is doctoral students and early-career researchers from partner universities, contributing as part of a longer programme of work co-supervised with our practitioners. The second is mid-career and senior practitioners — engineers, data scientists, consultants — who participate alongside their day jobs and contribute the kind of grounded judgement that academic teams alone rarely possess. The third is industry leaders and operators with deep domain expertise, who participate as problem owners and as critics of the work in progress.

Members who thrive in the network share a few characteristics. They are comfortable with the slower pace of careful research and equally comfortable with the impatience of production deployment. They write well and frequently. They take peer criticism as a contribution rather than as an attack. They follow through on commitments without being chased. And they hold themselves to the same standard of intellectual honesty whether the result is convenient or inconvenient.

Members who do not thrive — and we have several years of evidence on this — tend to share an opposite set of characteristics: an unwillingness to write up work in progress, a preference for spectacle over substance, and a difficulty in distinguishing between disagreement and disrespect. We are honest with applicants for whom we suspect the fit will be poor, and we have learned that this honesty saves everyone time later.

Membership is a real commitment. We expect members to contribute approximately a day a fortnight on average, with periods of more intensive involvement around active projects. We expect members to participate in the bi-weekly internal seminar where work in progress is presented and critiqued, either in person or asynchronously. We expect members to honour the publication and IP arrangements that govern any project they join, which are written down before the project begins. And we expect members to treat the network's other members — across institutions, geographies and seniorities — as colleagues whose time is as valuable as their own.

We also ask members to be transparent about competing commitments. Many of our members are full-time employees of organisations whose interests intersect with ours; the way to manage that well is to declare it openly and to recuse cleanly when it matters. We are equally transparent about the constraints that govern the work we can offer to any individual member, and we expect the conversation about constraints to be candid in both directions.

“Half a day a fortnight, done well and done reliably, is worth more to the network than three weeks of intense activity once a year.”

Members are offered access to the network's mentorship, with named senior researchers responsible for the development of each member's contribution to the Centre. Members are eligible for co-authorship on the publications that emerge from work they have meaningfully contributed to, on terms agreed before the work begins. Members can request access to the Centre's compute and data resources for approved projects, with allocations sized to the work and renewed at project review.

Members are invited to internal seminars, workshops and the occasional in-person convening that the Centre runs across the year. The convenings are deliberately small and are one of the most consistently appreciated aspects of membership; they are not events to be attended in numbers but conversations to be had in earnest.

Members in good standing also benefit from the Centre's relationships with industrial partners. Where appropriate, members are introduced to partner organisations whose interests align with their work, with introductions handled with the discretion that long-term relationships require. Several members have moved into senior in-house roles at partner organisations as a result of these introductions; we treat that as a positive rather than as a loss to the network.

Applications are read by at least two members of the Centre's leadership team and, where appropriate, by the principal researcher of the programme to which the application is most relevant. We are looking for evidence of three things. First, a serious technical foundation — though this need not be in machine learning specifically; many of our most valuable members come from adjacent disciplines. Second, a written voice that holds up to a paragraph of consideration; we read carefully and weigh what we read. Third, a clear sense of what the applicant would like to work on and why; vague applications are politely declined regardless of credentials.

We respond to every application within five working days, with one of three outcomes: an invitation to a longer conversation, a polite decline with a brief reason, or a request for a small piece of additional information that will let us decide. We do not maintain waiting lists; if we are not in a position to admit an applicant now, we say so honestly and invite them to apply again in the future.

We are happy to receive applications from candidates who do not match any of the standard profiles described above. The network is most valuable when it is composed of people who would not naturally have met one another, and the strongest applications are sometimes the ones that do not fit a template.