RAG (Retrieval Augmented Generation) pipelines bridge the gap between generative AI and real-world knowledge by combining retrieval and generation. This approach helps build reliable AI systems like chatbots, answering real-time queries or improving decision-making. RAG improves LLMs' generative capabilities by integrating real-time information from external sources, reducing hallucinations and inaccuracies. It enhances accuracy, scalability for domain-specific applications, adaptability to changing requirements, cost efficiency, and is less prone to hallucinations compared to traditional LLMs. RAG pipelines rely on data curation, efficient embedding storage, and a reliable data retrieval system to generate relevant output. They are susceptible to poor data quality, inefficient retrieval systems, inconsistent chunking, embedding overhead, scalability bottlenecks, and require continuous monitoring and improvement. Encord is a comprehensive data platform that simplifies dataset management, data curation, annotation, and evaluation, helping in the creation of RAG systems.