PIKE-RAG is a Retrieval Augmented Generation framework designed to overcome limitations of basic RAG systems in industrial settings. It extracts deep domain-specific knowledge from professional corpora and performs logical reasoning to augment LLM responses. The framework addresses challenges in meeting complex needs of real-world applications through specialized knowledge and rationale augmentation, published at ICML 2025.