§ about
AI engineer & researcher in Hamilton, New Zealand.
Right now, my main focus is the dissertation: building out a multi-agent system for data operations and all the AI engineering that comes with it. Think multi-agent design, getting LLMs to actually use tools. Before this, I spent 15 years building and shipping software across NZ, AU, the UK, Singapore, and the US. This includes multi-tenant SaaS, ML-powered products, cloud platforms, and plenty of time leading engineering teams.
Adematic — Independent AI Engineering
Adematic is my independent AI engineering practice, shipping production multi-LLM orchestration and agentic pipelines for SEO and MarTech clients including Visibility Labs and 180 Marketing. Models in active rotation: Anthropic Claude, OpenAI GPT, Google Gemini, Perplexity.
Multi-stage Rails pipeline that classifies and scores prospect URLs through Ahrefs metrics, eCommerce detection, GPT relevancy scoring, Gemini with URL Context tool for agentic page fetching, Gemini multimodal vision over screenshot captures, and Anthropic Batch API with Claude for multi-variant outreach.
Unified 9-tool SaaS for an agency managing 50+ clients. Multi-model AI Visibility tracking across ChatGPT / Claude / Gemini / Perplexity, embeddings + cosine similarity, RBAC, JWT auth, Playwright/Selenium scraping subsystem. Django + React.
Rails + Hotwire SaaS generating SEO long-form content via multi-step LLM prompt chains (LangChain, Claude, GPT-4) with Google Docs export and token-cost monitoring.
Pipeline that turns sitemaps into contextual internal-linking recommendations, combining LLM prompt engineering, async task orchestration (Django-Q, Celery), and content extraction (BeautifulSoup, trafilatura, Newspaper3k).
Bootyard — Engineering Lead
Led a 30-person engineering team at Bootyard, a SaaS consultancy delivering multi-tenant platforms, iOS and Android apps, and cloud-based systems.
Led platform and iOS development. Architected the multi-tenant SaaS that powered white-label rental rollouts for Selfridges, Matches, John Lewis, Flannels, and Rixo. Multi-tenant from day one; per-brand customisation, scoped data, isolated rental flows.
Joined as an early engineering hire. Contributed to platform engineering through a multi-million-dollar funding round and acquisition by TeamSnap in 2022. Built core platform from MVP through scaled production.
stack: Ruby on Rails · Hotwire/Turbo · PostgreSQL · Redis · ElasticSearch · React · AngularJS · Node.js · AWS · Heroku · Docker.
Mentored engineers, PMs, and QA in Agile, TDD, and legacy-code modernisation. The pattern repeated across every engagement: orchestration was the lever, not headcount. Small disciplined teams shipped faster than larger undisciplined ones, every time. It's the same lesson I'm now applying to multi-agent systems: the leverage is in how the agents coordinate, not how many you spin up.
Codetoki — Engineering Lead
Co-built Codetoki, a gamified developer-recruitment SaaS that assessed engineers through interactive coding challenges. Secured funding from the JFDI Asia Accelerator. Recognised by Microsoft, the Asian Development Bank, and the World Economic Forum for innovation and social impact.
Built in-browser compilers for Ruby, PHP, and JavaScript to evaluate developer skill and automate coding assessments at scale. Sandboxed execution, language-specific runtimes, and an evaluation harness for automated grading — a problem class that turns out to overlap meaningfully with the sandboxed-code-execution and tool-use surfaces of the dissertation system, a decade later.
stack: Ruby on Rails · MongoDB · Windows Azure · JavaScript · PHP · JRuby.
Master in Artificial Intelligence
GDip — Computer & Information Sciences
BS Computer Science
Three patterns recur across the fifteen years.
Greenfield system design.
Every role above involved building from scratch under real constraints — small team, deadline, scaling pressure.
Cross-cultural engineering teams.
Five countries, multiple engineering cultures. Shipped together by getting the orchestration right.
Engineering as research, research as engineering.
The most useful research insights came from production constraints; the most useful production decisions came from research-style rigor.
Open to conversations about applied multi-agent research.
See also: selected projects · writing.