LLM: Foundations&Applications

Catalog description

Provides a comprehensive exploration of Large Language Models (LLMs) and bridges the gap between foundational Natural Language Processing (NLP) and the frontiers of autonomous AI agents. Students trace the evolution of linguistic computation from classical word prediction models to the transformative self-attention mechanisms and multi-head architectures that power modern LLMs. Emphasizes advanced methodologies and practical design paradigms of LLM systems, including Retrieval-Augmented Generation (RAG), multimodal integration, and the development of LLM-as-Agent systems capable of tool use and multi-agent reasoning. Particular attention is given to the dual aspects of AI security, where students experiment with both adversarial attack techniques and defensive strategies. By treating LLMs as both a rigorous subject of study and a collaborative tool for improving code simplicity and algorithmic efficiency, the course prepares students to build, benchmark, and deploy responsible AI systems across diverse industrial applications. Prerequisite: INFO 501. Offered at least once every two years.