November 2026
I'm writing to apply for the Applied AI Architect, Industries role. What's actually driven my career across four decades and nine very different technical domains — equity derivatives, fixed income, commodities, FX, credit derivatives, hedge fund administration, and distributed network systems before any of that — isn't a résumé that wandered. It's imagination and taste in system design, and a compulsive need to keep learning something new. For the last fifteen of those years, that instinct found its best outlet as the sole architect of Axioma/Qontigo's flagship multi-asset portfolio and risk analytics platform (Deutsche Börse Group) — the kind of system institutional strategists trust with decisions they can't afford to get wrong, refreshed daily against years of market history. Before that, I spent nearly twenty-five years building similarly mission-critical systems for Morgan Stanley, Royal Bank of Scotland, Deutsche Bank, and Merrill Lynch.
Earlier this year I got to test that same instinct against the best tool I've ever had for it. I rebuilt the Axioma/Qontigo platform completely — a system that normally takes a team of three or four developers at least a year to get right, and budget only ever covered one person: me. I'd built the original solo, over several years, the first time. This time, working in direct, continuous collaboration with Claude as my engineering partner, I did it in three months — and shipped substantially more functionality than the original ever had. Not by cutting corners: the same architecture review and test-driven verification against live production data I've always held myself to, at a pace I didn't think was possible before. Claude didn't give me the taste for a better design — it gave that taste somewhere to move at full speed for the first time.
This moment is full of people getting impressive-looking results out of Claude, and it's tempting to think the tool explains all of it. It doesn't. Claude amplifies whatever judgment you bring to it — it doesn't manufacture judgment on its own. Mine took forty years and nine domains to build, long before Claude existed to borrow it from. That's the actual differentiator I'd bring to this role: not that I use Claude, but that I can tell, the instant it hands something back, whether it's actually right — architecturally sound, correctly scoped, safe to trust — or just confident-sounding. Most people experimenting with AI today are still building that instinct — which is exactly the line this role has to draw for customers, between adopting Claude seriously and merely experimenting with it. I spent four decades building that instinct the hard way, on systems where getting it wrong wasn't hypothetical.
That experience is why I want this specific role. Every engagement in my career has put me in the same seat: the person someone had to trust with something they couldn't verify themselves — a risk desk trusting margin logic they'd never audit line by line, a derivatives trading floor trusting a system to represent what they actually needed, investors trusting risk numbers they'd never see the code behind. This year, for the first time, I was on the other side of that relationship — the one who had to understand what Claude was actually doing well enough to trust it, push back on it when it was wrong, and vouch for the result to people relying on it. I know, directly and recently, both halves of what this role asks for: being a credible technical advisor to a skeptical, sophisticated audience, and understanding deeply enough how Claude actually behaves in production to be worth trusting on that subject.
This role appeals to me for the same reason my career has spanned so many different domains: it's built around constantly learning a new customer's problem well enough to design something genuinely good for it, not just correct. I'd welcome the chance to talk about what I think this role actually requires, and why four decades spent earning trust from people who couldn't check my work has prepared me well for a job built entirely around that same relationship, aimed at a different kind of complexity.
Thank you for your consideration.