How to work with us.
Three ways in. All three are vendor-neutral, outcome-bound and quietly engineered. No badges, no kickbacks, no theatre.
Pick the shape of the problem.
AI Feasibility Sprint
A fixed-scope diagnostic to validate whether AI is the right answer — and what the shortest path to value looks like.
- Data audit & feasibility report
- Prototype on your data
- Business case & ROI model
- Architecture blueprint
Production Engagement
End-to-end build of a production AI system, integrated with your stack, your data and your people.
- Model R&D & evaluation harness
- MLOps + monitoring
- Integrations & runbooks
- Operator training & handover
Embedded AI Team
An on-tap research & engineering pod that plugs into your roadmap. Vendor-neutral, outcome-bound.
- Senior ML + data + MLOps mix
- Quarterly outcome reviews
- Monthly research notes
- Pause & resume monthly
The non-negotiables.
Every model ships with a reproducible benchmark suite — no demoware.
Data, features, models, decisions — all versioned and queryable.
Drift, latency and outcome metrics on day one of production.
Runbooks and training so your team owns the system, not us.
Practical questions, candid answers.
Do you take equity instead of fees?+
Selectively, for venture-backed deep-tech where we genuinely believe in the thesis. Most engagements are fee-based.
Can you sign our paper?+
Yes. NDA, MSA, DPIA, model cards — we've been through it. We come with a default contract pack but happily redline yours.
Do you resell GPU capacity or licences?+
No. We are vendor-neutral by policy. We have no kickback arrangements with cloud or model vendors.
Can you work alongside our internal team?+
Most of our engagements are paired with internal engineers. The handover is built in, not an afterthought.
What happens after go-live?+
We offer optional retainers for monitoring, retraining and incident response. You're never trapped — exit terms are part of every contract.
How much does an AI project cost?+
Discovery sprints start at £18k for 2–3 weeks. Production engagements start at £85k for 8–16 weeks. Embedded teams start at £24k per month with a 3-month minimum.
How long does a typical AI engagement take?+
Feasibility: 2–3 weeks. Production build: 8–16 weeks. Embedded teams run continuously with monthly pause-and-resume.
Why engagement model is half the answer.
Most AI projects fail at the contract before they fail at the code. A fixed-price deliverable model rewards the vendor for finishing, not for being right; a time-and-materials model rewards staying. An IP arrangement that locks the client out of the model weights guarantees a future renegotiation under duress. A success fee tied to a metric the vendor controls tilts every modelling choice toward gaming that metric. None of these are abstract. We have inherited every one of them in salvage engagements, and the cost of unwinding them dwarfs the cost of having structured the relationship honestly at the start.
Our default contracts assume the opposite — that the client will own the weights, the data, the documentation and the right to leave; that the vendor's incentive is to make itself unnecessary; that the success measure is agreed in writing and computed independently. Those terms cost more to draft and they cost more to honour, and they are why our engagements end on schedule, with the operating team confidently running the system, more often than the industry average suggests is possible. The engagement model is not paperwork around the work — it is half of the work.
Start with a 30-minute call.
No slides, no sales pitch. We'll listen, then tell you honestly whether we're the right team.