For years, I engineered process systems at Australia's biggest organisations — cutting a 51-step company onboarding down to seven. Then I noticed small businesses missing leads because they couldn't afford the kind of automation those big companies take for granted. So I built Phoenix AI: an AI chat widget that runs locally, stays private, and works 24/7 so you don't have to.
Get in touch
The most concrete thing I did at ASIC was take a 51-step company onboarding process and cut it down to seven. Nobody else had sat down and asked if all those steps were necessary. So I did.
I did that kind of work for Westpac, ANZ, CBA, IAG, and others. Enterprise systems. Zero room for error. It was good work, but here's what stuck with me: every time a big organisation wanted to modernise, they had the budget, the teams, and the patience. Meanwhile, the local tradie running his trade from a ute was still fielding phone calls after hours, losing leads because nobody answered.
I thought: if I can automate complex enterprise workflows, why couldn't someone put an AI system on their website that answers questions when they're asleep?
The answer was it's too expensive or too complicated. So I built one myself. Phoenix AI is what happens when you take enterprise-grade automation and strip away everything a small business doesn't need — no API keys, no monthly fees, no data leaving Australia.
I've just released the PI Agent Harness bundle — a complete guide to building your local AI agent team. Includes a 440-page ebook, a 6-video series, ready-to-run configs, and all 13 agent definitions. Run a multi-agent AI coding harness entirely on your own hardware. No API costs. Full privacy.
Built in 10 minutes with the PI Agent Harness. Run pi.dev in a fully sandboxed Docker container — isolated from your filesystem, connected to your local model. See what a local AI agent team can ship.
I don't do consulting. I build tools and ship them. Here's what I've put into the world so far:
A single script tag embeds an AI chatbot on any website. It reads your site content, answers visitor questions in real-time, and works locally — no API keys or third-party data processing. Built for Australian businesses that want 24/7 customer support without sending their data overseas.
I've built and tested multi-agent systems running on consumer hardware — single GPUs, not cloud clusters. The idea is simple: each agent handles one job (researching, drafting, verifying), they coordinate through structured prompts, and the whole thing runs under your control. No vendor lock-in.
I run production AI workloads on an RTX 3090 — fine-tuning, RAG pipelines, inference. The Local AI Stack Blueprint documents everything I've learned: hardware choices that actually make sense for small teams, models that fit on a single GPU, and the mistakes I made so you don't have to.
Everything else I build feeds into this one product. Phoenix AI is the tool I wish existed when I was still consulting — a way for any Australian business to add 24/7 intelligent customer support without the enterprise price tag or the data-export compromise.
The AI chat widget that runs locally, answers your visitors 24/7, and keeps their data in Australia.
Build your local AI agent team — ebook, video series & configs. Zero API costs, full privacy.
Free demo project — built in 10 minutes with the PI Agent Harness. Sandboxed, isolated, fully local.
Technical write-ups, lessons learned, and practical guides from real projects.
How I distilled Claude 4.6 Opus chain-of-thought reasoning into a 24B code model on a single consumer GPU — seven bugs, a VLM weight extraction, and a transformers 5.x loading bug that took hours to find.