How we work

Ways to work together

Discovery sprint

A short, focused engagement that turns "AI could help here" into a concrete plan: what to build, what to buy, what to skip, and the order to do it in. It ends with a document your team can execute from, not a deck.

2–3 weeks

Founders and teams with an AI thesis and no agreed plan yet — especially when the disagreement is about scope rather than ambition.

Architecture & hardening

Design or repair the parts of an AI system that have to hold under real use: agent tool and permission boundaries, separation of trusted control from untrusted input, multi-provider failover, enforced structured outputs, and an evaluation harness so quality is a number rather than an impression.

3–6 weeks

Teams whose AI feature works in a demo and worries them in production — or who are about to give an agent real permissions.

Due-diligence review

An independent read of what is really under the hood before you invest, acquire, or scale: architecture, security posture, how safely the AI and agent components are built, and what breaks at scale. Reported plainly enough to be read by the people making the decision.

1–2 weeks

Investors sizing technical risk, and founders who would rather find the problems before someone else does.

Fractional CTO / hands-on build

Sometimes you need someone to actually build it. We embed with your team to architect and ship real product — Node/Postgres, Expo/React Native, WebRTC, Stripe, self-hosted infrastructure — from MVP through launch and into reliable operations.

Ongoing, reviewed quarterly

Teams without senior engineering leadership in-house, and founders who need the thing built rather than specified.

What happens after you get in touch

You write

Describe what you are building and where it is stuck. A paragraph is enough — there is no form to fill in beyond the one on the contact page, and no discovery call required to find out whether this is worth your time.

You get a real reply

Usually within one business day, and from the person who would do the work rather than from a scheduler. If none of the engagement shapes fits what you need, the reply says so — that is a faster answer than a call.

A conversation

One call, long enough to understand the system and the constraint you are actually up against. Anything that can be answered in writing is answered in writing instead.

A written scope

What the engagement covers, what it explicitly does not, what it costs, and how long it runs. Nothing starts until you have that in writing and have agreed to it.

The work

Progress you can see rather than a status report you have to trust: working software, or the document as it is being written. Scope changes get discussed when they come up, not discovered at the end.

Hand-over

You keep the artefacts and the reasoning behind them — the plan, the code, the evaluation harness, the operational knowledge. The engagement ending should not be the moment your team discovers what it does not know.