I architect agentic AI for regulated environments — the kind where shipping requires evidence, not enthusiasm.

Twenty-five years across platform engineering, MLOps, and applied AI. I've shipped Kubernetes-based MLOps at agency scale, designed autoscaling strategies for large-model training that delivered approximately A$1M per month in cloud cost savings, and led DevOps for financial-crime compliance programs at multi-team scale. Production AI infrastructure for organisations that don't get a second chance to fail audit.

Current focus: agentic AI architecture, multi-agent orchestration, evaluation harnesses for non-deterministic systems, AI FinOps, and the governance frameworks (PSPF, ISM, NIST) that take AI from "interesting pilot" to "approved for production."

What you'll find here: architecture writing for people who build AI in regulated industries — banking, government, utilities, healthcare, defence-adjacent. Specific patterns. Real trade-offs. No "ChatGPT changed everything" thought leadership.

I ship the code I write about: github.com/apur27/SuperCoach-VIA — end-to-end agentic AI reference implementation.

Monash MBA, Risk Management. Melbourne, Australia.