Early on, my work was the unglamorous operational layer: reconciling files that never quite matched, chasing confirmations across inboxes, and rebuilding the same report by hand every week. The pattern was always the same — the problem wasn't a lack of effort or talent. It was a lack of dependable systems.
So I started building them. I taught myself Python, JavaScript, APIs, and data pipelines, and began turning the processes I understood from the inside into automation that ran on its own — with validation, evidence, and a clear status a human could trust.
That operations-first perspective is still how I work. I don't start with a framework; I start with how the work actually happens: where information enters, where people lose time, which exceptions matter, and what a reviewer needs to sign off. The code comes after that's clear.
Today I focus on AI-assisted and agentic systems — but with the same discipline. AI is added only where it genuinely improves a workflow, never where it hides how a decision was made.