Full-Stack AI Engineer (Backend-Heavy)

We're open session, a design company building a design system tool for the future of work. Our mission is to help the world make the most of design and technology. We're building BOS, an AI platform that keeps every creative workflow on-brand. The front-end is near production, and now we're scaling and hardening the AI backend and building major new systems on top. We need an AI-native engineer who moves at the pace of a Claude Code-first team. **You'll own real, net-new systems, not maintenance:** - An AI **credits & metering** layer (usage tracking, pre-run cost preview, Stripe billing) - **Managed-agent orchestration** (routing, per-agent scope and observability) - **Multi-model routing** and **LLM observability** (token/latency/cost via OpenTelemetry) - Backend architecture for scale, RLS hardening, and CI/CD depth **Our stack:** Python 3.13 / FastAPI / asyncpg on Railway · Next.js 16 / React 19 / TypeScript on Vercel · PostgreSQL + pgvector (Supabase) · Anthropic SDK + LiteLLM · Voyage AI embeddings · OpenTelemetry · monorepo with git submodules. **Must have:** - Shipped **production async Python (FastAPI)** services - Real **AI engineering**: Anthropic SDK, RAG, embeddings/pgvector, streaming (SSE), agentic tool-use, not just API calls - **Claude Code power user**, current: fluent in the skills/subagents/hooks/MCP features that shipped through the second half of 2025 - Strong **PostgreSQL** (indexing, RLS, migrations) and **MCP** integration **Strong plus:** usage-metering/billing systems · enterprise auth (OAuth/SAML/OIDC/SSO) · Next.js/React + design-system fluency · prior AI products that scaled to enterprise web apps. **To apply:** Send 2-3 projects that scaled to production or enterprise (your specific role and the hard problem you solved), and a few sentences on how you actually use Claude Code day-to-day. We provision access to one of the most advanced internal Claude environments available. Growth is comp-based: rate increases on delivery. No equity.