If you Google my name, the algorithm will inevitably serve you a moderately dramatic British soap opera character. I am the other one.
I'm a Brisbane native currently attempting to survive London winters (and, lately, the summers), and for the last twelve years or so, I've been building products. Some of them worked brilliantly. Several of them cratered into the earth. I'm surprisingly okay with both outcomes.
My early career was consumed by a slightly unhealthy obsession with cyber security. As it turns out, spending your formative years trying to break into networks is phenomenal training for figuring out why your own production databases are mysteriously on fire at 3 AM.
I eventually did time in the enterprise mines at SEEK and Coles Group. It was highly educational in the exact same way that eating unseasoned broccoli is educational, but it showed me what actually makes large engineering organisations difficult. The politics, the change management, the sheer complexity of how hundreds of people try to build things together. That stuck with me. After enterprise, I led authentication at Linktree, built private equity infrastructure at Pactio (where the margin for error was a terrifying absolute zero) and chased down wildly obscure US carrier bugs while building a graph database at Blinq. The full list is over here.
I also have a terrible habit of building things. I founded OpenClub, a community management platform. As a founding engineer at AutoGrab, I helped build a predictive vehicle pricing engine that ingested over 20,000 data points a minute. We ended up outperforming Kelley Blue Book on residual price forecasting across Australia and New Zealand, which mostly proved that a few engineers with a massive data pipeline can comfortably ruin a traditional industry's day. After that, I built an AI assistant called Jamie. That project was mostly useful for teaching me exactly where AI creates actual, measurable value, and where it's simply a very expensive parlour trick.
Currently I'm the co-founder and CTO of Flowstate, backed by a16z Scout Fund, Haatch and others. The thesis fits in three words: tokens are labour. Every company now deploys human capital and token capital, and every system it owns was built for the first one. So AI spend turns up as the fastest-growing line on the P&L and the worst-measured one. Most teams I sit with can't tell you who spent what, on which model, for which project.
Flowstate puts both kinds of labour on one ledger. Every Claude call, every Copilot seat, every contractor day-rate tied to a person, a project and a budget. Then the harder half: forecast where it's heading and work out whether any of it was worth the money. The last piece is control. Set the budget up front, cap what a team or a model can burn through, and you don't learn about the overspend a quarter late.
It's the most ambitious thing I've tried to pull off. It's also the first time I've had a front-row seat to the bill.
When I'm not shipping, or apologising to my team for something I broke in staging, I'm either on a mountain somewhere, aggressively defending Australian coffee as the only objectively correct coffee on earth, or drinking AIX 2021 Coteaux d'Aix en Provence rosé. What I'm actually up to this month is on the now page.
Most of what I know I got from other people's writing. The ones I keep going back to are here.
I will not be taking questions on the wine choice.