Independent, Melbourne

Automation and AI for small and medium business

Take your business to the next level with the latest AI and cloud technologies.

I've worked in four-person startups and multi-billion-dollar Silicon Valley companies, in areas covering robotics, cloud operations, application security, and agentic tools. My passion is automating away the boring, repetitive work that belongs to computers, not humans. Modern AI tooling has opened up a new world of possible automations.

I use these tools myself to build software. That's why custom software is now cheaper than ever. I can ask my assistant to update sites I manage, like this one, in natural language, and it gets done in minutes: versioned, tested, and deployed. That's efficient.

Let the computers do the boring stuff, and let your people do more of what they're really passionate about.

What I do

Process automation: Quoting, scheduling, job records and the follow-ups between them, wired into the systems a business already uses.
Custom software: Tools written for the way a business actually runs, where nothing off the shelf fits.
Private AI: Models and search running on hardware you own, for work that cannot be handed to a public API.
Security and access: Credentials, permissions and network edges: the part that decides whether any of the rest is safe to run.
Systems and cloud: Linux, servers, deployment, and the monitoring that tells you when something has stopped.

Where the AI runs

For most of this work, the cloud is the right answer. A frontier model over an API costs a few dollars a month, runs on hardware someone else maintains, and improves without anyone having to touch it.

Local hardware earns its place when the data cannot leave the building: a law firm's client files, a clinic's patient records, contracts that are the product. The same class of model runs on a machine in the office, no request goes out, and the cost becomes equipment you own instead of a bill that changes every month.

It costs more up front than an API key. For some clients that is the whole of the decision, and for others it never comes up.

Fun stuff

Lex
An AI that runs on hardware in my study rather than through an API. It keeps its memory across conversations, holds its own set of tools, and has changed its own source code. The software behind lxg2it.com was written by Lex.
Bugs
An artificial life sandbox, built to be taught from. Every bug is a short string of DNA running as a small program: it moves, spends energy, eats, breeds with a neighbour and mutates, and the population evolves while you watch. For genetics that means descent, drift, selection and mutation in something you can follow for thousands of generations in minutes. For programming it means a machine whose instruction set fits on one screen, where changing a rule has a consequence you can see.
Humans
An inference API whose model is a person. Anthropic-compatible at the edge, priced at zero, settled in Bitcoin, and every request is read in full and answered by someone.
Stage
A viewer for OpenUSD, the scene format robotics and simulation teams describe their worlds in, and NVIDIA put 2,576 of their SimReady assets on Hugging Face under a licence that lets anyone use them. Every viewer draws those files and every viewer can draw them wrong: a part of the scene that never loaded, a surface that fell back to grey, a texture that did not arrive, all of it renders as something plausible. Stage draws the scene and puts the loader's own account of the load beside it, so the gap between what the file says and what reached the screen is visible. Pick one off the shelf, or drop your own folder in, where nothing leaves the machine.