AI Marketing Automation Architect

We're building an AI-driven marketing automation system for a real estate agency (Costa del Sol property market, primarily Dutch buyers). The system takes a new property listing as input, uses AI (Claude/Anthropic API) to analyze it, generates ad campaigns across Meta, Google Ads, Instagram, LinkedIn, email and WhatsApp, publishes them after human approval, tracks performance weekly, and feeds learnings back into future campaigns via a retrieval-based memory layer. We already have a complete architecture document (data flow, tool choices, API list, database schema direction, learning-loop design) — you won't be designing this from scratch, you'll be implementing and owning it, with room to challenge/improve our design decisions where your experience says otherwise. Tech stack we've specified: - n8n (self-hosted) for orchestration - Supabase (Postgres + pgvector) for data + learning layer - Anthropic API (Claude) for analysis, copywriting, and performance reasoning - Meta Marketing API + Graph API (campaigns, Instagram, WhatsApp Business) - Google Ads API - Image/video generation tools (Imagen/Flux/Creatomate) — open to your recommendations here We're looking for someone who can own the build end-to-end, not just execute isolated tasks. You'll work from our architecture doc, propose adjustments where relevant, and ideally continue as an ongoing technical partner for maintenance (API changes, new features) after the initial build. Phase 1 (fixed-price discovery, ~1-2 weeks): - Review our architecture document, flag risks/gaps - Set up Supabase schema (multi-tenant-ready from day one) - Set up n8n instance (self-hosted) with basic ingestion workflow (listing URL/upload → structured data) - Deliver a short technical plan for phases 2-4 Phase 2+ (ongoing, hourly or monthly retainer): - AI analysis + content generation workflows (Claude API integration) - Meta Marketing API + Google Ads API integration (including navigating their App Review processes) - Weekly performance sync + reporting workflows - Learning-loop implementation (vector store, RAG into generation prompts) - Basic approval dashboard (can start as Retool, evolve later) - Ongoing maintenance as APIs change Requirements: - Proven experience with n8n (show us a workflow you've built, ideally something with branching logic, not just linear zaps) - Experience integrating Meta Marketing API and/or Google Ads API — please describe a project where you did this, including how you handled the App Review process - Experience with LLM API integration (Anthropic, OpenAI, or similar) for structured output / tool use, not just chatbot wrappers - Comfortable working from a written architecture spec and communicating clearly in English (or Dutch) — we need clear written updates, not just code - Bonus: experience with Supabase, vector databases / RAG, or previous real estate / marketing automation projects How to apply: Please include: - A link or description of a previous n8n project with non-trivial logic - A short note on how you'd approach the Meta Marketing API App Review process, since this is often the biggest time-risk in this kind of project We will not grant access to production ad accounts initially — you'll work in a sandbox/test ad account until the integration is proven, then get limited-role (non-admin) access to production.