QA Engineer

You'll take complete ownership of quality for assigned areas of Cityflo's rider and operations products — working shoulder-to-shoulder with engineers and PMs, not behind them. That means deriving high-coverage test plans from PRDs and Figma, automating the cases that matter across Android and iOS (cross-device, varied networks, OEM quirks, the store release lifecycle), and debugging both QA and prod down to a clean repro. You'll use LLMs daily to generate cases, surface edge cases, draft automation, and triage crashes — and you'll know exactly when to verify the model against the app, the code, and the spec. This is a startup quality role: you scope ruthlessly, you push back, and the bug that double-charges a commuter is your job to find and refuse to let ship.

What we look for

• 1–3 years owning quality for real shipped product, mobile-first (Android + iOS): cross-device, cross-OS, flaky networks, OEM quirks, the app-store release lifecycle
• Derives high-coverage test plans from PRDs and Figma — E2E, backend, regression, exploratory — and surfaces the gaps where the spec is silent rather than guessing
• Hands-on with at least one of Appium / Maestro / Playwright / Espresso / XCUITest; comfortable writing, running, and triaging an automated suite
• Fluent with LLMs (Claude / ChatGPT / Cursor) to generate cases, draft automation, and triage logs — and knows when to verify the AI against the app, the code, and the spec
• Debugs both QA and prod: reads crashes, logs, and traces to a root cause and a clean, minimal repro
• Startup-style ownership: thrives in ambiguity, scopes ruthlessly, and pushes back when something is wrong — including on a payment bug someone would rather see buried

The assignment

We point you at Cityflo's real staging rider web app (a live React SPA, phone + OTP login) and hand you a short PRD for "Monthly Pass + Auto-Renew." Drive your agent to turn the PRD into a crisp test plan, automate the handful of cases that actually matter (Playwright preferred), run them against staging, triage failures, and file clear bug reports with repro steps and defensible severities. Staging payments are sandboxed and free, so the full purchase → payment → auto-renew funnel is safe to exercise end to end — but you authenticate with your own number, never use a real card, and test respectfully. The signal isn't volume: it's whether you verify the live app against the spec instead of trusting the PRD or your AI, find where reality and the spec diverge, and scope well under time pressure. AI is expected throughout. Aim for 2–4 focused hours.

2–4 focused hours. Going much past that is a signal we'd rather not see — a small, sharp slice with excellent reasoning beats a giant suite you can't defend in the debrief. You do not need to fix anything; you're QA here, not the implementer.