Academics

  /  

Courses

  /  

Descriptions

IEMS 490: Theory and Algorithms for LLMs

VIEW ALL COURSE TIMES AND SESSIONS

Prerequisites

Graduate standing, or permission of instructor

Description

Generative AI is transforming how we learn, create, and build software—and this course provides a rigorous, hands-on introduction to the theory and algorithms behind today’s Large Language Models (LLMs). You’ll learn how transformer-based models are trained and deployed, then dive into the techniques that power real-world genAI applications: prompt engineering (including advanced and algorithmic approaches), vector databases and retrieval-augmented generation (RAG), and post-training with reinforcement learning methods such as RLHF and direct preference optimization (DPO). The course also covers large reasoning models, in-context learning, and efficient adaptation via fine-tuning and low-rank methods (LoRA), with a look at emerging foundation models beyond text (e.g., time-series). The course assignments require use of code assistants to develop implementations for a variety of use cases and topics. Practical implementations by using modern software techniques and deep understanding of the theory of LLMs are two main goals of the course.