Python Backend Developer for GDPR-Compliant RAG/MCP Knowledge Server

Project goal: Build a GDPR-compliant knowledge server, hosted on a Hetzner GEX44 GPU server in Germany. The server forms a frontend-agnostic core with an internal retrieval API, to which a lightweight MCP server and a later custom interface attach as parallel clients. Main tasks: Development of an internal retrieval API as the central core (parallel client attachment, not in series) Integration of Qdrant as vector database with a per-tenant collection structure Integration of Ollama running Qwen3 14B locally Setup of the database infrastructure with PostgreSQL Configuration of the entire environment via Docker Compose Implementation of an MCP server over Streamable HTTP (no SSE) with OAuth 2.0 Ensuring German-language answer quality, measured against a provided test question set Requirements: Demonstrable RAG/LLM project experience (concrete project) Solid Python backend expertise Experience with vector databases (Qdrant, Weaviate or pgvector) Docker skills MCP experience a plus Fluent German or English DACH/EU timezone Expected deliverables: Working Docker Compose setup Runbook Test suite Handover session Code in the client's Git repository; IP ownership transfers to the client from the first commit How to apply — please include: A concrete RAG/LLM project from your own work A brief take (2–3 sentences) on how you would separate the internal retrieval API from the MCP server, so that the MCP server and a later custom UI attach as parallel clients rather than in series Your estimate of the answer latency for Qwen3 14B on a Hetzner GPU server Your fixed price for Phase 1 Optional but welcome: a one-page architecture sketch Working model: DACH/EU timezone