You build a prototype that works perfectly in a demo, but the moment it reaches production, it breaks. This is where most AI projects fail. In this talk, Joanna will explain why this happens, share the best practices for building stable and scalable AI systems, and show how an effective evaluation setup can keep quality consistent.

Retrieval-Augmented Generation (RAG) has become a popular method to address this issue, augmenting LLMs with an external knowledge base. However, implementing RAG introduces distinct challenges. In this presentation, Joanna will share practical insights into the challenges encountered while implementing RAG systems, alongside strategies for overcoming them. You'll be equipped with the tools and methodologies needed to navigate these challenges successfully.