AI products that survive contact with production.

Most AI work dies between the demo and the deployment. We run generation, marking and evaluation pipelines in production on our own platform, at a scale where quality has to be measured rather than assumed. We build the unglamorous half: evaluation, guardrails, cost control and the human review step that keeps the output defensible.

Our own work in this sector

We have shipped this, not just pitched it.

Ilmino question pipeline

Generation, verification and quality gating of exam questions at a scale of tens of thousands, with independent re-solving and sampled review.

AI marking

Typed and handwritten student work marked against a mark scheme, with staff able to override.

What usually goes wrong

The problems we are normally called in to fix.

The demo works and the product does not

A prompt that looks impressive once behaves unpredictably across ten thousand real inputs. Without evaluation you find out from customers.

Nobody knows what it costs to run

Token spend scales with usage and quietly becomes the largest line in the bill. It has to be designed for, not discovered.

You cannot tell whether a change made it better

Without an evaluation set, every prompt change is a guess and every regression is invisible until somebody complains.

Output nobody will stand behind

In regulated or high-stakes work, a confident wrong answer is worse than no answer. There has to be a human step and an audit trail.

What we build

Software and marketing for ai & saas.

AI product development

The whole product, not the model call. Auth, billing, admin, usage limits and the interface people actually work in.

Agents and workflow automation

Multi step processes that read, decide and act across your systems, with the failure paths designed rather than hoped for.

Retrieval over your own content

Answers grounded in your documents and data, with citations, so the output can be checked rather than trusted blindly.

Evaluation pipelines

Scored test sets, independent verification and sampled review, so a prompt or model change is a measurement rather than a gamble.

Cost and latency control

Model routing, caching and batching, so unit economics work at ten thousand users rather than only at ten.

B2B SaaS platforms

Multi-tenant applications with organisations, roles, subscriptions and the reporting a buyer expects before signing.

Services we bring to this sector
  • AI product development
  • AI and automation
  • Custom software development
  • Website design and development
  • Mobile apps
  • Hosting and support
What it is for

The commercial outcome we are aiming at.

Questions we get asked

Straight answers.

Which models do you build on?

Whichever fits the job, and usually more than one. Routing a cheap model for easy cases and an expensive one for hard cases is often the difference between viable and unaffordable unit economics.

Can you work on an existing AI product rather than start fresh?

Yes, and it is often where we add most. Teams frequently have a working prototype and no evaluation, no cost visibility and no guardrails, which is exactly what stops it shipping.

How do you stop it producing wrong answers?

You cannot stop it entirely, so you design for it. Grounding in your own data, evaluation sets that catch regressions, confidence thresholds, and a human review step wherever being wrong is expensive.

Do we own the model and the code?

You own the code, the prompts, the evaluation sets and the data. We do not hold your product hostage.

Tell us what you are trying to automate, and we will tell you if AI is the right tool.

A couple of lines is enough to start. You will get a straight answer on what it needs, what it costs and how long it takes.

Reply within one working day

Tell us what you need building and who it is for. You will hear back from the person who would do the work.

Discuss Your Project