Architecture
Turning ambiguous requirements into systems, standards and decisions that teams can actually build against.
Selected work
Things worth showing.
A small selection of the problems I spend time on and the engineering ideas that have come out of them.
Current focus
Engineering reliable systems around probabilistic models.
I am exploring where LLM reasoning is genuinely useful, where deterministic controls are better, and how to verify outcomes without trusting model confidence.
I write about the parts that generalise: the design choices, failure modes and engineering lessons that remain useful outside the original problem.
Read the latest articleAreas of work
Architecture, AI-assisted engineering and assurance.
AI-assisted engineering
Using LLMs for design, synthesis and problem solving without pretending they are deterministic software.
Verification
Moving stable rules into repeatable checks and using observable evidence to decide whether work is done.