This paper describes the system developed by a team for the TUM practical course Human-Centered Computing: applications in natural language processing, network science, machine learning, and AI for the SemEval MTRAG. Our approach addresses the challenges of multi-turn retrieval-augmented generation (RAG) by combining context-aware query rewriting with a dense retrieval strategy. We employ a pipeline that cleanses noisy corpora and utilizes dense OpenAI embeddings via Milvus for robust retrieval, and leverages Gemini 2.5 flash family of models for standalone query generation and final response synthesis. Our system demonstrates the effectiveness of integrating high-precision retrieval with fact-based generation across diverse domains.