Subject-matter experts. Diligent practitioners.

Most AI consultancies are dominated by one of three professional cultures — the academic research culture, the software-engineering culture, or the management-consulting culture — and many of their failure modes can be traced back to that imbalance. We deliberately recruit and develop across all three, with a fourth specialism in domain expertise drawn from the industries we serve most often. The mix is not always frictionless, but it produces work that is at once intellectually serious, operationally credible and genuinely useful to clients.

Within research, we cover machine learning, computer vision, natural-language processing, optimisation, statistics and the classical knowledge-based traditions. Within engineering, we cover data engineering, MLOps, distributed systems, full-stack application development, security and infrastructure. Within consultancy, we cover strategy, programme management, governance and change. Within domain expertise, we have practitioners who have spent significant parts of their careers inside manufacturing, retail, financial services, healthcare, telecommunications, energy and the public sector.

Most engagements draw from at least three of these specialisms; some draw from all four. The shape of the team is matched to the shape of the problem rather than the other way around.

Our hiring bar is high and our hiring process is correspondingly slow. We look for technical depth, of course; we look harder for clear thinking under pressure, written communication that holds up to a fortnight rather than a paragraph, and the kind of intellectual humility that survives contact with a difficult client. The candidates we hire are usually overqualified for the role they join in; the candidates we decline are often technically excellent on paper.

We say no to candidates who cannot describe their previous work in terms of the decisions it changed; we say no to candidates whose written work shows more confidence than the underlying argument supports; and we say no to candidates whose references describe them as brilliant but difficult to work with. A small firm that depends on collaboration cannot absorb the cost of brilliance that does not collaborate, and we have learned this the hard way more than once.

We hire deliberately across geographies, backgrounds and career paths. Diversity of background is, in our experience, the single largest contributor to the quality of the questions a team asks; the quality of the questions, in turn, is the largest contributor to the quality of the work.

“Brilliance that does not collaborate is a liability dressed as an asset.”

Our development model rests on three commitments. First, every practitioner has a written development plan, refreshed twice a year, with explicit objectives and the support required to meet them. Second, every practitioner spends a meaningful share of their time — typically a day a fortnight — on work that is not directly billable: research, internal tools, mentoring, writing. Third, every practitioner is expected to teach as well as learn; the seniority ladder is partly built around how much of the team's collective capability you have helped to build.

We do not separate the research track from the engineering track from the consulting track in the way larger firms tend to. Many of our most senior practitioners move fluidly between deep technical work, client-facing leadership and internal capability building, and we structure compensation and progression to reward that fluidity rather than penalise it.

The result is a team that grows people who are unusually hard to lose. Our voluntary attrition is low, and the practitioners who do leave us tend to leave for roles we are happy to have helped them prepare for — including, in several cases, senior in-house roles at clients we continue to work with.

We are mostly remote, mostly hybrid in client engagements, and entirely intentional about both. Remote work is supported by clear written norms — when to write, when to meet, when to escalate — and by tooling that makes asynchronous collaboration genuinely productive rather than a polite fiction. We invest in regular in-person time per team and across the firm, because the relationships those weeks build pay dividends across the months that follow.

We hold ourselves to a high standard of written communication. Architectural decisions are written down before they are made; client conversations are captured in shared notes within hours; major opinions live in documents long enough that they can be challenged and revised. This discipline is occasionally tedious and is, on net, the single largest contributor to the quality of our work.

Disagreement is welcomed and structured. We expect practitioners to argue technical positions hard, to lose those arguments without resentment when the evidence demands it, and to commit to the agreed direction once a decision is made. The middle of those three is the one most often forgotten elsewhere; it is the one we treat as non-negotiable.

We hire continuously rather than to fill specific seats, and we are particularly interested in mid-career and senior practitioners with deep applied experience. The roles that recur most often are: senior machine-learning engineers; data engineers comfortable across modern cloud stacks; computer-vision specialists with industrial deployment experience; MLOps engineers; consultants with applied AI delivery experience; and research scientists with a track record of moving research into production.

We are happy to hear from people who do not see their exact background reflected in the list above; the surest route is a written description of the kind of work you want to do next, the constraints under which you would like to do it, and the contribution you believe you could make. We respond to every credible enquiry within a fortnight.