I run an AI consulting and implementation firm that builds intelligent automation systems for business clients across multiple industries. I am looking for a skilled individual developer to be our long-term technical partner — building and maintaining client systems as we grow. Individual freelancers only. No agencies. No teams. ─────────────────────────────── Required skills n8n — complex architecture Sub-workflows, error handling, webhook triggers, modular design. Not basic linear flows. Claude API or OpenAI — advanced prompt engineering Structured data extraction, categorisation accuracy, consistent JSON output. Not chatbot work. AI agent architecture Per-client context profiles, structured memory systems, automated learning from human corrections. REST API — write-access operations OAuth 2.0, multi-client credential management, write operations not just read. Google Workspace Drive, Sheets, Gmail — including structured data management and dynamic context loading. ─────────────────────────────── Strong bonus QuickBooks Online or Xero write-access API — not just connecting, actually posting entries via API with OAuth and sandbox testing. Multi-tenant system design — separate builds per client, isolated credentials and data. Cross-industry automation experience — built for more than one vertical. ─────────────────────────────── This engagement Each client gets a fresh build — no reused systems, no templates. You build clean, documented, production-ready workflows every time. This is not a one-off project. If you want a single payment and move on — this is not for you. If you want consistent work, clear briefs, on-time payment, and a retainer that grows as we add clients — read on. ─────────────────────────────── To apply — send all four 01 A Loom walkthrough of the most complex n8n workflow you have built — not a screenshot, a live walkthrough showing you understand what you built and why. 02 Explain specifically how you would call Claude API from inside n8n to extract structured data from a document and return a consistent JSON output. 03 Describe how you would build a per-client memory system in n8n — where each client has their own context profile that Claude uses and that updates automatically from reviewer corrections. 04 Your project rate for a 150–200 hour build and your expected monthly retainer for ongoing maintenance. ─────────────────────────────── Applications without a Loom walkthrough or answers to questions 02 and 03 will not be reviewed.