Nine months running engineering at a retail-intelligence startup — and the case I point to for how I work now: using AI to build faster and more efficiently.
Its customers were some of the biggest names selling in Brazil — Ambev, GPA, Haleon, Corteva and Grupo Boticário among them — so the platform’s pricing data carried real commercial weight.
I joined as interim CTO with a short handover from the departing CTO, then took over the roadmap: prioritising the team’s technical decisions while staying hands-on in the TypeScript and React codebase. One of the projects was migrating a product database from MongoDB to ClickHouse. We were partnering with another company that provided the conversational, ChatGPT-like AI layer — our side was to feed it the data, so customers could ask questions of numbers they used to export to spreadsheets. That needed queries far faster than MongoDB could serve; a columnar SQL database like ClickHouse was a much better fit for the analytical, read-heavy access it demanded. I did the preliminary work — adapting the Python APIs and translating the MongoDB aggregation pipelines into ClickHouse queries, with AI assisting along the way — before handing it to a senior backend engineer to take over and finish.
Two more things shipped in parallel: a major upgrade of the React Native mobile app, and a Reseller Prospecting Portal for Boticário — Brazil’s biggest beauty brand — an MVP built primarily with AI that mined the existing user base to identify and qualify potential resellers.
It’s the case I show for owning a roadmap and a team without leaving the code — and for treating AI tooling as a way to deliver, not a talking point.