11-766: Large Language Model Applications (CMU, Spring 2026)
by Carnegie Mellon University (Fernando Diaz, Daphne Ippolito)
CMU's graduate course on turning LLMs into real systems: prompting vs fine-tuning, RAG and deep research, agents, LLM evaluation, attacks and deployment.
Overview
11-766 Large Language Model Applications ran at Carnegie Mellon University from 13 January to 23 April 2026, taught by Fernando Diaz and Daphne Ippolito with six teaching assistants, and its full site (schedule, lecture slides, syllabus and homework) is public at cmu-llms.org and versioned on GitHub. The syllabus frames the course around practical questions: does an application need a general-purpose model or custom training, and what evaluation and infrastructure does it require? The 25 topic lectures run in four arcs. Foundations: origins of LLMs, natural language understanding versus generation, the science of prompting, deciding when to fine-tune and fine-tuning efficiently, and learning representations and embeddings. Retrieval: three lectures on storing and retrieving knowledge, retrieval-augmented generation and deep research, plus a dedicated deep-research session. Interactive systems: task-oriented dialogue; tool use, chitchat, personas and companionship; writing and ideation assistants; LLMs for evaluation through synthetic data, simulation and automatic evaluation; multi-agent systems; and code-writing assistants. Risk and breadth: harms caused by LLM applications, attacking LLMs and LLM applications, non-English languages and non-American cultures, world models, biology, music generation, numbers, robots and embodied AI, and deployment. Three homeworks (30% of the grade) mix written and coding questions: Prompting, Finetuning, Retrieval and Embeddings, delivered as two Jupyter notebooks, an annotated dataset and a LaTeX report; Chatbots, Agents and Assistants; and Language Models beyond Natural Language. A 40% final project explores one application in depth. The materials are slides-only; no lecture recordings are linked on the schedule, so self-learners work from slides, readings and the homework.
At a Glance
- Topic
- RAG
- Level
- Advanced
- Format
- Course
- Cost
- Free
- Duration
- Spring 2026 semester (13 Jan - 23 Apr): 25 topic lectures as slides plus 3 multi-week homeworks; budget ~8 weeks at ~6h/week self-paced
- Provider
- Carnegie Mellon University (Fernando Diaz, Daphne Ippolito)
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Decide when prompting is enough and when an application justifies fine-tuning, and how to fine-tune efficiently
- ✓Learn representations and embeddings and use them to store and retrieve knowledge for LLM applications
- ✓Build retrieval-augmented generation pipelines and extend them into multi-step deep research systems
- ✓Design task-oriented dialogue systems, tool-using assistants, personas and companion chatbots for real users
- ✓Use LLMs for evaluation through synthetic data generation, user simulation and automatic evaluation methods
- ✓Understand how multi-agent systems and code-writing assistants are structured, including a guest lecture on coding agents
- ✓Analyze the harms caused by LLM applications and the attacks used against LLMs and LLM-powered apps
- ✓Apply language models beyond natural language to images, music, biology, numbers and embodied robots
Highlights
- •Four consecutive sessions on retrieval, RAG and deep research (29 Jan - 10 Feb 2026), unusually deep coverage for a general LLM course
- •Downloadable homework packages with starter Jupyter notebooks, PDF write-ups and LaTeX templates, so the coding work can be done outside CMU
- •Breadth rarely found elsewhere: guest lectures on code-writing assistants, world models, LLMs for biology (Prof. Lei Li), music generation (Prof. Chris Donahue) and embodied AI
- •Taught by Daphne Ippolito, who also teaches CMU's 11-667 LLMs: Methods and Applications, making 11-766 the applications-focused follow-on to that fundamentals course
- •The whole course website is versioned on GitHub (cmu-llms-class/cmu-llm-class-website-2026), so slides and assignments stay available after the term ends
Who It’s For
Best For
- ✓ML engineers who know transformer fundamentals and want a structured tour of LLM application patterns
- ✓Engineers choosing between prompting, fine-tuning and RAG for a product feature and wanting the research behind each
- ✓Graduate students and researchers looking for a slide-and-reading-driven survey of LLM applications beyond chatbots
Prerequisites
- •A prior course in language-model fundamentals; CMU lists 11-667, 11-711, 10-423 or an equivalent class
- •Python and Jupyter notebooks for the coding portions of the three homework assignments
- •Comfort reading recent ML research papers, since slides are the only lecture material
FAQ
What is 11-766: Large Language Model Applications (CMU, Spring 2026)?
11-766 Large Language Model Applications is Carnegie Mellon's Spring 2026 graduate course, taught by Fernando Diaz and Daphne Ippolito, for engineers who already understand how language models work and now want to apply them. It covers prompting versus fine-tuning, embeddings, RAG and deep research, chatbots and agents, LLM-based evaluation, attacks and deployment, so you can design and justify an LLM system for a real application.
Is 11-766: Large Language Model Applications (CMU, Spring 2026) free?
11-766: Large Language Model Applications (CMU, Spring 2026) is free to access.
What level is 11-766: Large Language Model Applications (CMU, Spring 2026) for?
11-766: Large Language Model Applications (CMU, Spring 2026) is aimed at a advanced audience. Recommended background: A prior course in language-model fundamentals; CMU lists 11-667, 11-711, 10-423 or an equivalent class, Python and Jupyter notebooks for the coding portions of the three homework assignments, Comfort reading recent ML research papers, since slides are the only lecture material.
How long does 11-766: Large Language Model Applications (CMU, Spring 2026) take?
Expect roughly Spring 2026 semester (13 Jan - 23 Apr): 25 topic lectures as slides plus 3 multi-week homeworks; budget ~8 weeks at ~6h/week self-paced. Most learners work through it at their own pace.
What will I learn from 11-766: Large Language Model Applications (CMU, Spring 2026)?
You'll learn: Decide when prompting is enough and when an application justifies fine-tuning, and how to fine-tune efficiently; Learn representations and embeddings and use them to store and retrieve knowledge for LLM applications; Build retrieval-augmented generation pipelines and extend them into multi-step deep research systems; Design task-oriented dialogue systems, tool-using assistants, personas and companion chatbots for real users; Use LLMs for evaluation through synthetic data generation, user simulation and automatic evaluation methods; Understand how multi-agent systems and code-writing assistants are structured, including a guest lecture on coding agents; Analyze the harms caused by LLM applications and the attacks used against LLMs and LLM-powered apps; Apply language models beyond natural language to images, music, biology, numbers and embodied robots.
Topics
Sources
This page was written from 6 sources, 2 on domains other than cmu-llms.org.