Senior Python / AI Engineer to build AI platform for real estate agencies (RAG, API integrations)

I'm building an AI-powered SaaS platform for real estate agencies, starting with the UAE market. The goal is to cut the manual work agents do every day with developer brochures, payment plans, listings and leads. Full requirements are in the attached PDF. Please read it before applying. What the platform does: • Turns unstructured developer documents (PDF brochures, payment plans, floor plans, price lists) into a structured, searchable property database • Lets agents search, filter and compare projects and units with exact, source-backed data (price, size, payment plan, handover date, service charges) • Generates client-ready materials: unit comparisons, offer summaries, follow-up messages in English and Arabic • Qualifies incoming leads and matches them to suitable units • Syncs with CRMs and listing portals via API • Offers an AI assistant for agents and an embeddable widget for agency websites MVP scope: • Multi-tenant backend on FastAPI + PostgreSQL • Document ingestion and structured data extraction from messy PDFs and tables • Hybrid search: structured filters + vector search (pgvector or Qdrant, your recommendation) • LLM layer with RAG, source citations and strict validation for prices and numbers • REST API for integrations, first CRM connector • Eval set to measure extraction and answer accuracy • Docker, CI/CD, AWS deploy Requirements: • 5+ years Python in production • At least one shipped LLM/RAG product with real users • Strong experience with PDF and table data extraction • API integrations with third-party systems (CRM, webhooks) • You make architecture decisions and explain trade-offs • React/Next.js for a simple admin panel is a strong plus • 3-4 hours overlap with GST Process: paid test task (4-6 hours), then MVP in ~6-8 weeks, then long-term work if we're a good fit. How to apply: • Start your proposal with "payment plan" • Answer: how would you make sure the platform never shows a wrong price or payment plan? • Send a rough estimate for the MVP scope from the attached PDF: hours per main block (ingestion and extraction, search, LLM layer, API and CRM connector, deploy) and the total • Record a short Loom (2-5 minutes): introduce yourself and walk through one LLM/RAG project you built - the architecture, the main problem you solved and what you would do differently today Proposals without a Loom and an estimate will not be reviewed.