About

Twenty-five years in technology across New Zealand's public sector and private industry, at home and abroad. For the last four of them I have worked hands-on with AI: building open-source multi-agent systems, writing down engineering practice for agent-assisted delivery, and coaching people from AI-curious through to experienced engineers.

The same principle runs through this site: I would rather show something working than describe it. Technology is changing faster than at any point in our careers, and the teams that do well will be the ones whose people are ready first.

I lead the Ministry of Social Development's 24-person Architecture practice, one of the largest architecture functions in the country, and I am accountable for architecture delivery across MSD's Technology group.

I designed the practice's blended operating model after the 2024 IT restructure, bringing Enterprise, Domain and Solutions Architects into cross-functional teams aligned to the business. I have taken the practice through five change processes, from founding the unified practice to a significant headcount reduction, while keeping morale and delivery up.

Most of what I am proud of there is about people rather than architecture: rewriting every Architecture position description, designing a gamified progression framework called the Architect's Role Compendium, and a coaching framework to go with it. As contract manager I formed MSD's Architecture Panel, shortlisting thirty RFP responses to a five-member panel under a $9.6m three-year budget.

I am the lead voice for AI enablement across the practice and advise the Director of Emerging Technology. MSD recognised that work with its 2023 "AI Whisperer" and "Creative Talent" awards.

Before the practice, I spent a decade as an architect across the full lifecycle: as Infrastructure Architect at MSD, and before that as a Solutions Architect at Fujitsu, in trusted-advisor roles for government and commercial clients.

I led Fujitsu's Citrix consultancy practice in New Zealand and helped architect GoDaaS, New Zealand's first multi-tenanted Desktop-as-a-Service platform, including its self-service portal for automated approval and deployment. After the November 2016 earthquake, GoDaaS restored Statistics NZ's secure digital workplace within four days. In 2017 Citrix recognised the platform as one of the world's most advanced multi-tenanted architectures.

The practice from that period that matters most to me is decision records. I introduced ADRs as standard practice at MSD with executive endorsement and built the decision log that made governance workable. I then presented MSD's guardrails approach to the Government Enterprise Architecture Group, which led to collaboration with Stats NZ, Te Whatu Ora and DIA. The same discipline governs every project on this site.

I co-founded the New Zealand chapter of the Agentics Foundation in 2025 and grew it into a nationwide community for practical, responsible agentic engineering.

In practice that means monthly hands-on AI Hackerspace sessions and in-person meetups, coaching practitioners from AI-curious users through to experienced engineers, and building open-source multi-agent systems in the open with safety and quality controls built in. Unsorry and Sovereign Engine both came out of that community rather than from me alone.

Community work is the hardest test of whether a method works. A method that only works when I run it is not a method.

I live in Wellington with my wife and our two young children. Time outdoors with them is the part of the week I protect most. I meditate daily and I climb.

I also once co-founded a craft liqueur company in a Lower Hutt garage. I ran its copywriting, compliance, operations and manufacturing alongside a full-time job, won a trophy for Best Liqueur in Aotearoa, and sold the company in 2021. It taught me more about shipping than any technology role has.

Underneath all of it, what I care about is human-centred AI, and whether the people around me are ready for what is coming. Usually that shows up as turning something into a quest.

Every word on this site, and every ordering and grouping, lives in an Iris collection rather than in the code. The pages, their sections, the navigation, the six themes and the projects in them are all elements in that model. Only the design system lives in the repository.

When the site builds, it reads the collection through Iris's anonymous export, checks it, reads the star counts from GitHub, and renders static pages from the result. If Iris is unavailable, the build uses the last export that passed its checks, and the summary below says so.

The summary is rendered from the same export as everything else, so it cannot disagree with the site. It shows what the model holds, when it was exported, and the raw data of the element that draws it.