About

I trained as an architect, then in building science at Columbia on the spatial intelligence and interaction track. Architecture teaches one habit above all: take an under-specified brief, find the constraints that actually bind, and design something that holds under them. The medium changed — physical space, then data, then software systems — but I have been working on the same problem the whole way: space, relationships, constraints and behaviour inside complex systems.

Two jobs turned that into what I build now. Inside a garment manufacturer I ran product development and overseas business, watched a planning team lose hours every week to manual follow-up across more than a dozen departments, and wrote the order-tracking tool that fixed it. At a consulting firm I turned hundreds of thousands of public records into a framework growth teams could use to decide where to go next — and automated the research behind it with Python and language models.

The quoting system was the natural next step: a workflow I knew from the inside, documents I had read for years, and a problem — unstructured inputs that have to become exactly-right numbers — that is a near-perfect test of how to build with models responsibly.

How I work with AI

Most of the code in my recent systems was written by AI coding agents under my direction. The architecture, the acceptance criteria and the quality gates are mine; so is the responsibility when something is wrong. I treat that division the same way I treat it inside the products: decide what may be generated, what must be verified, and what a person has to own.