Selected work

Real tools, built for real businesses, in use today.

Each of these started with a business owner and a problem that was eating their time. I scoped it, built it with AI-assisted tools, and got it into their hands. Here's how they work and why they're built the way they are.

Marketing operations

AI Marketing Agent

A managing director with a distinctive voice but no time to post. This tool drafts LinkedIn posts in his voice, lets him approve them in minutes, and publishes on schedule — no marketing hire required.

Voice-matched AI Approval cockpit Auto-publishing
Read the case study
AI Marketing Agent approval cockpit showing a draft LinkedIn post ready for review
Field operations

Field Report Engine

A pest control technician was losing evenings to report writing. This tool turns rough site notes and photos into a branded, client-ready PDF in minutes, with the technician approving every word before it goes out.

AI rewrite PDF generation Mobile-first
Read the case study
Field Report Engine reports dashboard showing draft and finalised site visit reports

How I build

A business operator who builds with AI.

I ran a catering company for four and a half years before I built software. That's not a footnote — it's the whole approach. I start from the operational problem, not the technology, and I use AI-assisted tools to build the solution fast without losing control of the quality.

01 — Discover

Discover

Sit with the owner and find where the time actually goes. The real version, not the org chart.

02 — Validate

Validate

Agree the problem and what success looks like before a line of anything gets built.

03 — Build

Build

Build fast with AI-assisted tools, in small iterations the client sees early and often.

04 — Verify

Verify

Watch it used for real, then harden the rough edges. Shipped and used beats polished and theoretical.

Validate before building. The most expensive code is the code that solved the wrong problem.

Match the tool to the trigger. Scheduled automation for repeating tasks, AI for the human-initiated ones.

Anything client-facing gets a human gate. A few seconds of oversight is what makes automation trustworthy enough to use.

Knowing when not to build is part of the job. One of last year's best decisions was a feature I cancelled.

The stack

The tools I build with.

I'm fluent in a modern AI-assisted toolchain and pick the right tool for each part of the job rather than forcing everything through one.

Lovable

Front-end and cloud scaffolding — the interface the client actually uses.

Claude Code

Logic-heavy work: the functions and rules that run behind the interface.

Claude API

The AI that drafts, rewrites, and matches voice inside each tool.

Make

Automation and publishing — wiring the pieces together and out to other systems.

Supabase

The database, storage, and auth that hold everything each tool remembers.

Have a process that's eating your time?

Start with a free audit: a structured session that maps your workflows and shows where automation will actually make a difference.

Book a free audit