Large Language Models (LLMs) represent one of the most significant advances in artificial intelligence during the last decade, embodying what industry experts increasingly recognize as a technological inflection point comparable to the arrival of the internet. This chapter provides an examination of LLM architecture and functioning, with particular emphasis on their application in technical domains such as petroleum refining. The chapter begins with an analysis of the Transformer architecture that forms the foundation of modern LLMs, explaining the self-attention mechanism, positional encoding, and multi-layer processing. It then explores the training methodologies that enable these models to develop their remarkable capabilities, including pretraining on massive text corpora and fine-tuning for specialized applications. The chapter addresses the critical epistemological limitations inherent in working with LLMs, acknowledging that these systems operate as complex statistical models whose inner workings cannot be fully traced or predicted deterministically. Finally, it examines the practical applications of LLMs in petroleum refining and petrochemical processing, identifying both opportunities and limitations for industrial implementation. This technical foundation provides essential context for understanding how prompt engineering techniques can be effectively applied to complex industrial challenges while maintaining appropriate awareness of system limitations.