Curiosity first.
Proof always.
I’m an applied AI product builder and the founder of Auromind, based in Peterborough. I came to software by building tools for my own work, then kept going.
I learn by building
and checking.
My background spans project administration, frontline work in the probation service and trading. Each of them rewards the same habit: deciding on evidence, keeping clear records and noticing when something doesn’t add up. Software became the way I put that habit to work.
In Leaf Edge, I followed a trade from its individual executions to a review I could trust. In MindCap, I made small speech and language models compete on benchmarks I specified, then judged them inside the app. In Samplify, I listened for the clip that didn’t match its selection. Twin Shadows pushed me to carry a visual idea across a whole application.
Along the way I built my own system for running AI coding agents safely: isolated copies of the code, checks that read what actually changed, and nothing merged without my approval. I now bring that way of working to founders and small teams with an idea or a process that needs to hold up.
Read the project storiesAI in the workflow.
Ownership in the delivery.
I use AI coding agents throughout development. My role is to define the problem, set the constraints, direct the implementation, review the evidence and decide what ships.
I’m open about that approach. Agents, established libraries and third-party models do a lot of the typing. I take responsibility for the result.
For client work, that means a written scope, working previews, checks against the agreed behaviour and a clear handover. For an unfamiliar system, I start by understanding it before committing to a price or timetable.
A clear next step.
A useful first version with a scope we both understand.
Work you can check.
Working previews, the tests behind them and direct communication as the build progresses.
An honest handover.
Source, setup, known limits and a defined support arrangement.