About

I work on AI controls architecture in financial services: how institutions make AI systems accountable once they move beyond answering questions and begin using tools, changing records, and acting in the world.

I began in IT audit, then moved into banking data science before consulting. That path made the central problem clearer. Model capability is rarely the limiting factor; the harder work is turning probabilistic behaviour into controlled workflows with evidence, ownership, intervention, and recovery.

These essays are written for people who already own part of that problem: AI leads, model-risk specialists, controls architects, technology-risk leaders, and engineers building agent systems. They assume familiarity with the field and focus on mechanisms, evidence, and operating consequences.

The twenty selected essays develop that argument across controls architecture, assurance, agent systems, and work after AI. The archive preserves the experiments and working notes behind it.

My private lab tests these ideas in software. It uses cell biology as a design constraint for permissions, compartmentation, lifecycle, and failure handling, because naming the mechanism often reveals gaps before the abstraction hardens.