I started in performance marketing, running campaigns and A/B tests. But I kept finding the real bottleneck wasn’t the campaign. It was the plumbing underneath: data that didn’t flow, tools that didn’t talk, and expensive software that did only eighty percent of what we needed.

So I started building the missing parts myself.

A doctor directory that replaced a paid third party tool and opened a new revenue line through partner listings. A signup pipeline that moved sensitive health data off shared hosting into a pseudonymised system on my own server. Partner reporting that took days each month, automated down to almost nothing.

This isn’t engineering for its own sake. Custom makes sense when the alternative is expensive, inflexible, or requires handing over data you shouldn’t. Otherwise, buy the tool. Knowing which situation you’re in is most of the job.

Right now I'm building AI into that same layer: agent workflows that handle the operations work that used to need a person, and campaign infrastructure that runs on scripts and APIs instead of manual clicking.

Away from the keyboard I solve Sudoku, which is the same pleasure as debugging a pipeline but with fewer stakeholders.