CPA (Australia) and finance engineer. I build the controls into the data.
Most people building finance systems come from the technology side. I came from the business side, so I know what a number means before I write the query, and what an auditor will ask before they ask it.
Accountant, builder, coach. Curious about how work gets done, compassionate with the people doing it, and I empower them to change it.
A balance sheet reconciliation that pulls its own balance, open items and evidence, signs itself off when it's low risk and unchanged, and reopens itself when someone posts a late journal.
- Driven by a GL master. Ownership, audit risk, thresholds and misuse rules are data, not code.
- Gated by a certified query and a trial balance tie-out every close.
- Nine planted cases, from a missing invoice to a stale sign-off, each one caught. GitHub Actions proves it on every push.
The working methods I've written for Claude. The main one, session-close, closes an AI working session so the documents, decisions and next steps actually land.
- Start from the standard. A readiness checklist built from public frameworks (COSO, SOX and PCAOB, COBIT, DAMA), then the GL master as the foundation everything else reads from.
- Certify the queries, not the snapshots. Prove the data once, and no one witnesses or retypes a balance again.
- No roll-forward. Each rec is one living file that pulls its own evidence, and every number is backed by a source document.
- Bring the close to the accountant. The system flags what's missing and proposes the journals, so people spend their time on the high-risk accounts.
- AI prepares, a person signs off. AI works under its own identity, never under a person's name.
- A control that catches nothing proves nothing. The checks I build are proven against planted failures.
I'm not a coder, and I don't pretend to be. I direct AI to write the code, and I check every piece against standards I set: naming conventions, validation rules, reconciliation back to the source system, and documentation written as I build, not after. The accountant in me doesn't trust a number until it ties out, and I hold AI output to the same standard.
- I built a governed general ledger data platform in Databricks, from the raw SAP layer up, with a formal certification path every object goes through before anyone relies on it.
- At GetYourGuide I delivered an SAP BRIM cutover in a module I'd never touched, on six weeks' notice: 18 entities, 40 currencies, three million transactions a day. That was before AI.
- As Head of Global General Ledger I led 15 people across 18 entities, as a manager of managers.
- Fifteen years inside manufacturing businesses, from the shopfloor up: inventory, production, costing and procurement.
- I coach teams to adopt AI and own the change themselves, rather than making it for them.
Industries: SaaS, FMCG, agriculture, manufacturing, aerospace and logistics.