Melbourne, Australia
AI & Agentic Engineer
Introduction
Kintsugi is the Japanese art of mending broken ceramics with seams of gold — the repair becomes the most beautiful part. That's how I build software: take the brittle thing, the manual process, the system straining at its limits, and rebuild it with something stronger and a little luminous.
These days that means agentic AI — production LLM tooling, MCP servers, sub-agents and multi-agent pipelines — as Dev Lead of the Agentforce squad at RACV, on a foundation of years shipping on demanding enterprise platforms.
LLM tooling, MCP servers, sub-agents, evals — agents that do real work, not demos.
Internal tooling analysing thousands of customer interactions a day, in production.
Leading delivery on Salesforce's agent platform — Apex, LWC, Experience Cloud underneath.
MuleSoft, DataWeave, REST — a bias for making brittle things fast and resilient.
Projects

A roguelite card-duel game — custom rules engine, animated board, multiplayer, shipped to the web.
Play it →
Local-first LLMs and multi-agent systems — MCP servers, sub-agents, provenance-tracked asset pipelines, voice interfaces.
On GitHub →
Always another seam to lay — from 24-hour game jams to original AI experiments. I never stop building.
Say hello →In the detail
This is a card from The Academy, rendered live in a single WebGL shader — a crystal frame with cut facets and gold trim, a parchment face rasterised to a texture, and a foil layer that answers the light.
The foil isn't random sparkle. Hotspots are pinned to the frame's natural flare points — corners rake with streak-crosses, edge-midpoints bloom with diffuse spectra — and each one flashes white-hot only when the moving light truly aligns with it, the way a real foil card behaves in your hand.
Details matter. This one took an evening of sliders, argument, and gold.
See it in the game →Experience
Leading development on Salesforce's AI agent platform — technical direction, design review, and hands-on build with the squad.
Built internal AI tooling on Azure for analysing customer-contact data at scale.
LWC in a customer-facing Experience Cloud community on the high-volume new-business journey.
Salesforce and MuleSoft; took one integration from 30+ minutes to ~70 seconds.