• 5+ years shipping product end-to-end (ideation to deployment) that customers actually feel
• Deep in backend distributed systems, with a strong grasp of mobile app ecosystems (native plus our shared KMM and WebView layers)
• Designs APIs, data layers, and platform infra that other teams build on; raises developer velocity org-wide
• Reasons from fundamentals and first principles over familiarity with any one stack
• Understands LLMs and designs automations for support, developer productivity, and operational efficiency, including the trust boundaries they need
• Has UX and accessibility sense, and uses AI dev tools to contain complexity and tech debt, not just go faster

The Assignment

A Cityflo support team is drowning: ~1,800 tickets a day in one undifferentiated queue, routine refunds and 'where's my bus' sitting next to safety incidents that get found too late. We hand you a messy real-world dataset (support tickets plus the trip and booking records behind them) and you decide what to build. Write a one-page product memo, then build the smallest useful end-to-end slice that puts an LLM where it actually earns its keep (triage, drafting, reconciling claims against trip data), with a real UI a human can use and a decision trail someone could later defend to a rider, to finance, or to a regulator. Build it like a platform seam, not a one-off. Some of the text in this dataset was written by people trying to game whatever reads it; design for that.

Up to a weekend. We expect 8–12 focused hours of real work; going much past that is a signal we'd rather not see. Deliberate scoping is part of what we grade: a sharp slice of the data, handled honestly, beats a shallow pass over all of it.