Aden Mann

I build applied AI systems for work where being wrong is expensive: agents, evaluation loops and end-to-end workflows, built to hold up in production and run without someone watching them. My work typically sits in financial operations, where a wrong number is a real cost.

Before working in tech I spent over a decade flying Australian Army helicopters in special operations aviation. That’s work where being wrong is catastrophic, and it taught me how human-machine systems actually fail: rarely as an obvious breakdown, more often as a confident, wrong reading that nobody catches in time.

I then spent four years at Immutable, a US$2.5B web3 company. I ran a ten-figure digital-asset treasury under tight governance, in one of the least-regulated corners of finance, where the discipline built into the system was the only thing holding it together. After that I built the company’s AI function from the ground up as Head of AI and Automation: an AI fluency framework mapped across more than a hundred people, 86% Claude adoption under security-approved guardrails and a company-wide build day that shipped 105 working agents into production. I built the guardrails as well, including a traffic-light governance framework that let the company move quickly without a serious incident, and a retrieval agent that took routine legal and compliance queries off the team.

These days I build many systems myself and hold them to the same bar. I care about building from first principles and not iterating on things that shouldn’t even exist in the first place.

I have an MBA with Excellence from AGSM at UNSW via scholarship, and I also founded Rotor Arcade and Valour Coffee Co.

If you’re working on something in this space and want to compare notes, let me know.