Forward-deployed AI engineer & fractional CTO.

The 25-year arc

Three platform shifts, same problem each time: how do you actually run this in production? Virtualisation in the 2000s. Cloud in the 2010s. Generative AI now. Same job, three eras of new toys.

Mainframe (1995–2000)

Bank of Scotland, systems programmer. The era where you learned how production actually worked — change-control windows that closed at 05:00 sharp, run-books written in pencil on bound paper, the operations bridge where every problem ended up.

It's a mainframe.

Everything I now know about production discipline I learned from people who'd been running the platform of the day since before the web existed. It turns out that's the same discipline GenAI needs now.

Loudcloud / Opsware (2002–2004)

The original "AI-ops" company, back when it was just ops. UK Cabinet Office account; helped roll out direct.gov.uk and the Iraq Dossier infrastructure. Marc Andreessen and Ben Horowitz's first ops play — and a clean view of what happens when "automation" meets a sovereign customer with sovereign timelines.

VMware (2004–2009)

Built the VMware Operational Framework while consulting at HSBC, Barclays, RBS, BNP Paribas. The methodology is still in use at all four banks more than fifteen years later. That framework is the lineage hook for what VOF is now: a written-down, shippable opinion about how you run the platform of the day.

Cloud era (2009–2019)

Cisco UCS. VCE Office of the CTO. Founding CTO at Canopy Cloud — a $0 → $280M JV between Atos, EMC, and VMware. Microsoft Hybrid Cloud, where the team won the internal "Best VMware migration to Microsoft" award. Then Cloudsoft, where I took the practice from AWS Standard to Advanced Consulting Partner in eighteen months.

AI era (2022– )

Cybersecurity AI and automation at SJULTRA, where Injectionator started as an in-engagement tool and is now a project of its own. Now: forward-deployed GenAI engineer building VOF and its four flagships, Ambassador at The Centre for GenAIOps.

Most teams have a working RAG pipeline. What they don't have is an answer to who uses it, when, and how — and that's the question your CISO and your CFO are about to ask.