Senior Full-Stack Engineer for a B2B Marketplace (FastAPI/Supabase/React, Claude Code, Linear)

I'm building a B2B wholesale marketplace connecting nurseries/growers with landscapers who need to find who has specific plants in stock, right now and near them. It's discovery-only (no on-platform transactions). The product has moved past prototype: the walking skeleton (auth, deploy pipeline) is live on staging and production, and the seller upload pipeline including a two-pass AI-assisted matching system for reconciling plant identities across sellers is substantially built and has been validated against real production data from active growers. We will need to test this more together. I'm looking for a senior engineer to take ownership of engineering going forward so I can focus fully on product, design, and go-to-market. What you'll own: Full engineering architecture, end to end: +Data model and service boundaries +Search/matching architecture +Deployment and infrastructure decisions +Testing strategy +Code review (all PRs, including ones generated with Claude Code) You'll work independently within this scope. I'm not reviewing code line by line. I review at the decision level: what shipped, what changed, what risk you flagged, and why you made the call you made. What I own: Product, design, prioritization, and anything a customer sees or experiences. I write specs with acceptance criteria, explicit out-of-scope boundaries, and a visual (via Claude Design) for anything user-facing. When a technical question changes what a user experiences, I want to be looped in on the tradeoff in plain terms before you build it not asked to approve a stack choice. Stack: +Backend: FastAPI, modular monolith (features → accounts/catalog → core) +Database/Auth: Supabase (Postgres + Auth, JWKS/ES256) +Hosting: Railway (API), Vercel (frontend) +Frontend: React / Vite / Tailwind / shadcn +Monitoring: Sentry +AI: Claude Opus 5 / other models for adjudication tasks in the matching pipeline How we work: +Claude Code is required - this is how you'll execute in the repo. I want an AI-native workflow, not code written entirely by hand. +Linear for issue tracking - I write tickets with acceptance criteria and out-of-scope notes; you ask clarifying questions there before starting. +GitHub for PRs - one clear PR per unit of work, with a short note on what decision was made and why. +ADRs for any architectural decision - this is the gate before implementation, not a retroactive writeup. +Failure-mode table required before implementation on anything nontrivial: what happens if this happens twice, what happens if it never happens, what happens if two people do it at once. +Weekly sync (video), async the rest of the time. No daily standups although optional. +"Show me before you act" — no merges or destructive actions bundled into the same step as showing me the result. Required experience: +Production experience with FastAPI (or comparable Python backend framework) and Postgres at meaningful scale - not just tutorial-level CRUD apps +Comfortable owning architecture decisions independently and documenting them clearly for a non-technical product owner +Experience with AI-assisted coding workflows (Claude Code or similar agentic tools) - you should already be using something like this, not learning it on this project +Search/matching or data-quality pipeline experience is a strong plus, given the matching system involved +Startup or small-team experience where you've had to make judgment calls without a large eng org backing you up +Has built a product for the 0 to 1 stage of company building Timeline: +Targeting functional completeness by November 2026. Current pace and scope discipline make this feasible for early user onboarding (already lined up).