In today's AI landscape, organisations increasingly demand systems that give responses grounded in real, current data—especially in non-English contexts, where many large models fall short. Our team at InoGen AI set out to build a Retrieval-Augmented Generation (RAG) system for our Dutch clients, using open-source components, to offer high quality, transparency, and real utility. In this article I'll share the architecture, additions, design choices, challenges, and benefits of our approach, using sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 for embedding and Llama 3.1 70b as our generation model.