Product Engineer — AI-Native SaaS (Cloudflare + Claude Code)

Product Engineer — AI-Native SaaS (Cloudflare + Claude Code)** We build a modern CRM platform for the roofing industry — web, mobile (iOS/Android), telephony, payments, document AI, in-app AI assistant — and we build it differently: **AI agents write the code, and you direct them.** We've spent months building an orchestration harness on Claude Code: parallel agent lanes, automated verification and design-critic agents, deploy gates, written playbooks. You'll operate and improve it. You won't type much code — but you need a solid working understanding of what the agents build, because you own whether it works and is worth shipping. **The stack you'll be directing work in:** - Cloudflare Workers (a multi-worker fleet), Durable Objects, D1, R2, Queues, Postgres, wrangler multi-env deploys - Vue 3 + Quasar web frontend; Ionic/Capacitor mobile (iOS + Android) - Integrations: Stripe, telephony/SMS, OAuth, push notifications - Anthropic API — agentic features inside the product itself **Hard requirements:** - General code and infrastructure literacy — you understand how workers, databases, deploys, and auth fit together, can follow what the agents built, read an error message, and tell when something looks wrong. You don't need to be a hand-coding senior engineer; you need to never be lost. - Hands-on familiarity with the Cloudflare ecosystem (Workers, wrangler, deploys) — or proof you learn platforms like this fast - Daily-driver fluency directing Claude Code (or equivalent) at scale: subagents, parallel lanes, automated verification — actual shipped output, not chat prompting - **Elite UI/UX judgment — non-negotiable.** You design and ship polished, dense, professional interfaces — light/dark, mobile/desktop — no designer on staff. Portfolio must show interfaces *you* designed. - Evidence-driven skepticism: you never accept an agent's "done" — you demand the screenshot, the live walk, the passing test, and check it yourself - Exceptional written communication — your instructions are the engineering; vague writing ships broken software - Relentless self-learner: you'll inherit a large system and learn it from its docs, and you'll keep those docs current as you go **To apply, send:** 1. Something real you've shipped by directing AI agents — link it, and tell us how you drove and verified it 2. 2–3 examples of product/UI work where your taste shaped the result — live URLs preferred 3. A short note, in your own words: how do you verify that something an AI built actually works? First-round screen is a paid, timeboxed exercise on our stack: we hand you our docs and a real feature request; you direct Claude to ship it. We watch how you read, instruct, and verify — and whether you catch what the agents get wrong.