Full-Stack Engineer — Liminal Spaces

Full-Stack Engineer — Liminal Spaces Liminalspaces.ai is becoming the trusted operating layer between complex property transactions and the households navigating them. Upload inspection reports and disclosures; get AI-powered, evidence-linked risk scores, cost-to-cure estimates, permit history, and negotiation framing — in minutes. We're a small founding team moving fast with AI-assisted development, and we're launching soon. We're hiring our first senior engineer to be the technical conscience of the codebase — the person who owns code quality and stands up the engineering guardrails an early-stage product needs. Enormous ownership, direct founder access, real product impact from week one. What you'll own Code health & technical stewardship (why this role exists): critically review a fast-moving, AI-assisted codebase; spot correctness bugs, security holes, and scaling traps before they bite; and stand up the guardrails an early-stage product needs — automated tests, CI, evaluation harnesses, error monitoring, and security/RLS review. You'll own the hard, interesting core of the engine and the eval work that keeps findings accurate as inputs vary. The AI document-intelligence engine (our core IP): a Python / FastAPI service on Railway that turns inspection PDFs and disclosures into structured, source-cited findings, risk scores, cost estimates, and permit history. Heavy on Claude (Opus / Sonnet / Haiku), PDF and vector parsing, with a strong emphasis on grounding, citation, and eval-driven accuracy across varied real-world document types. Supabase backend: Postgres with real row-level security, Auth, Storage, Realtime, and a suite of Deno edge functions. RLS here is a trust boundary, not a checkbox — you'll own that sensitive household and disclosure data stays correctly isolated. Frontend: React 18 + TypeScript on Vite — a web dashboard plus a Capacitor-based mobile/PWA app. Monetization & integrations: Stripe subscriptions, credits, and access gating; plus permits, pricing, and Google (geocoding / Drive). You'll thrive here if you Are senior — you've been the person who owns code quality on a team — and can review AI-generated code critically, not just produce more of it. Have shipped production apps on Supabase and can write and debug non-trivial RLS policies. Are fluent in Python for real LLM pipeline work — structured output, grounding/citations, evals — not just calling an API. (We're a Claude shop.) Think in async: queues, callbacks, idempotency, job recovery, stuck-state handling. Make pragmatic calls about what to harden now vs. later, and care about the user, not just the code. Bonus points FastAPI, prompt/eval engineering, document-AI / OCR, Stripe webhooks, mobile/PWA (Capacitor). Logistics Location: Remote — overlap with US Pacific hours preferred Type: Full-time / contract-to-hire Comp: TBD Tiny team, direct founder access, real ownership from day one. To apply Send your GitHub or portfolio plus one short paragraph on either the gnarliest RLS or async bug you've shipped a fix for, or a time you inherited a messy/AI-generated codebase and what you triaged first. No cover-letter theater