Career

“I don’t want to keep building that from scratch each time.”

Every platform I’ve built started with that thought. Most of any solution is boilerplate, so I build it once, properly, and then create a factory to rinse and repeat.

Ecommerce sites, event apps, health and safety checklists, and now software itself. Four startups, four acquisitions, thirty years of building, and I’ve been hands-on in every one.

startups built and sold
4
building for the web
Since 1996
event apps shipped
1,500+
invoices a month through my AI pipeline
20k

Netwise

Founding engineer · Acquired by the NAEA

My first job out of university: sole developer on an early online property portal.

  • Sole developer from the first line of code
  • Interactive property maps built from Ordnance Survey data
  • Acquired by the National Association of Estate Agents

When I joined, the company was a CEO and a set of designs. The idea: estate agents upload their properties and they appear on a map. Think an early Domain or realestate.com.au.

There were no mapping APIs in 1996, so I worked out how to slice up Ordnance Survey map data and build an interactive property map myself.

JavaScript · Perl · IIS · MS SQL

Acorn Ecommerce

Founding CTO · Acquired by Pindar PLC

My first factory: a drag-and-drop website builder that let a team of three run around 150 online shops.

  • Around 150 ecommerce sites on one platform
  • £50m+ in orders processed
  • A team of three
  • Acquired by Pindar PLC in 2007

We started out building ecommerce sites for clients. After a few by hand, I built the platform instead: a drag-and-drop website builder with order processing, stock management and integrations with all the major payment processors.

Under the hood, the front end was a pure JavaScript app talking to an API backend, years before anyone called that a single-page app.

JavaScript · VB.NET · IIS · MS SQL

Genie Connect

Founding CTO · Acquired by Lanyon

My second factory: a no-code app builder that shipped 1,500+ event apps across five platforms.

  • 1,500+ event apps shipped
  • Five platforms driven from one configuration
  • Apps for around $10k where competitors charged $70–80k
  • 30+ staff; I led a dev team of 12

The iPhone had just landed, and my co-founder had an idea: mobile apps for events. I built the first one by hand. Then I thought: I’ll be buggered if I’m going to do that again each time.

So I built the app factory. Organisers dragged in widgets such as exhibitors and sessions, picked colours, uploaded a spreadsheet or connected their event system, and out came an app.

One configuration drove five clients: iOS, Android, Windows Phone, BlackBerry and mobile web, each with its own specialist-led codebase. I stayed the main architect, and kept writing code, throughout.

Dojo · Objective-C · Java · GCP

Career break

After the Lanyon sale I took time out to travel, and landed in Sydney in 2016. I’ve been here ever since.

Contract development

Developer

Freelance development work in Sydney while settling into Australia.

OpsBase

Founding CTO · Acquired by Ideagen

My third factory: a checklist maker for hospitality, steered through COVID to an acquisition.

  • Hundreds of clients, thousands of checklists
  • Dev team in Sydney, sales and marketing in the UK
  • Survived COVID all but wiping out its market
  • Acquired by Ideagen in 2021

Same instinct, new market: a configurable checklist builder (checkboxes, photo uploads, text fields) that walked hospitality staff through things like closing-up procedures and health and safety checks.

Then COVID all but shut hospitality down. We cut back hard, from six developers to one plus me, kept the platform running for the customers who remained, and sold the business to Ideagen in 2021.

React · React Native · GraphQL · Node.js · AWS

Non-profit SaaS

Founder

Started to build my next SaaS product: a financial planning tool for non-profits. Then AI changed everything.

The Lab

Independent R&D

From early 2023, alongside client work: three generations of my own agent runtime, ThoughtBeat, Silas and Iteri.

Everything I learned in the lab went straight into client systems, and everything clients needed fed back into the lab.

Read about the Lab →

LeapForward

AI architect (contract) · Handed over to the client’s team

An AI care assistant with strong guardrails and a secure hand-off to human support workers.

  • Chatbot on the OpenAI API, inside our own architecture
  • Strong guardrails throughout
  • Secure human-in-the-loop hand-off
  • Handed over to the client’s team in Singapore

OpenAI API · AWS

Republic Plc

AI architect (contract, via Delta Tech) · 20,000 invoices a month in production

Enterprise invoice processing with AI, chosen over competing solutions from other vendors.

  • 20,000 invoices a month in production
  • Vision and OCR combined for near-perfect extraction
  • Multiple LLMs with automatic failover
  • Chosen over competing vendors’ solutions

Republic wanted to know how AI could help their business. Invoice processing was the perfect first target.

I built a pipeline that turned each PDF into images, ran OCR over them, and gave the model both, so it could read the page and the text together. Accuracy was close to 100%.

LangChain was fine for the prototype. For enterprise scale I moved orchestration to AWS Step Functions, with automatic failover between LLM providers.

AWS Step Functions · multiple LLMs · vision · OCR

Squiz

Agentic AI architect (contract) · Delivered ahead of time

A standards-compliant A2A framework, so teams across the business could build agents that work together.

  • Fully compliant A2A agent framework, 100% AWS serverless
  • Dynamic agent cards and tool manifests for cross-team development
  • Assessed AWS Bedrock Agents for production fit
  • Delivered ahead of time

A2A · AWS serverless · Bedrock · Nova

Fusion5

Principal AI Architect · Built the AI Software Factory prototype

Designed and built an AI Software Factory: describe the problem, get a well-architected Azure solution.

  • Working end-to-end prototype
  • Security, pipelines and monitoring built in by default
  • AI writes the business logic; the factory supplies everything else

Fusion5 delivers Microsoft solutions for clients. I proposed an agentic Software Factory to change how that work gets built, and built a working prototype on Azure.

The business wasn’t ready to move away from its traditional consultancy model to adopt it, but the factory proved the idea.

Azure · Cosmos DB · infrastructure as code · agentic AI

How the Software Factory works →

What’s next

I’m looking for my next role designing and building AI systems: hands-on, close to the product, with room to build the factory.

Get in touch →

Techie #1. Chief architect. The product guy.

My title has usually been CTO, but in a startup that means whatever the business needs. For me it has always meant:

Always hands-on

Even leading a team of twelve, I was the main architect and still writing code. I lead from within.

From vague to working

“We need gamification.” My favourite part of the job is turning an idea like that into something implementable that fits the shape of the whole product.

Generalise, then build

When a customer asks for X, I ask what X is an instance of, and extend the platform so it covers X and the next five requests too.

Builder, not administrator

I build and lead dev teams. Timesheets and HR are best left to someone who enjoys them.

Toolkit

AI & agents
OpenAI · Anthropic · Gemini · Bedrock · LangChain / LangGraph · CrewAI · A2A · MCP
Languages
TypeScript / JavaScript · Python · Java · Perl
Front end
React · React Native · Angular / Dojo · Three.js
Cloud
AWS (serverless, Step Functions, Amplify) · Azure · GCP · Docker · Terraform / CloudFormation
Data
SQL · DynamoDB · Aurora · Cosmos DB · Elasticsearch · vector databases
APIs
REST · GraphQL · JSON-RPC · OpenAPI · JSON Schema