Courses and labs
Work through LLM concepts in guided programming exercises.
Use course notebooks to build a working prototype for assessment.
PROGRAM IN DEVELOPMENT
LLM
Learning Lab
Learn to build and test
LLM applications.
Work through model fundamentals in practical labs, seminars or mentored research. Learn to check an application's accuracy, cost and risks as well as write the code.
Ask about the programFor students, educators and research collaborators.
What you'll study
-
01
How language models work
Understand tokens, attention and Transformers, from pre-training and post-training to inference.
-
02
Prompts and source material
Write prompts and use retrieval-augmented generation (RAG) to answer questions with sources you can check.
-
03
Adapting models and using tools
Study supervised adaptation and preference learning. Set permissions for the tools an agent uses.
-
04
Testing an application
Use test sets and human review to find errors. Measure latency and cost.
PROJECT EXAMPLES
- A reading assistant that cites its sources.
- A cultural-heritage learning app, with an evaluation of its results.
TEACHING & RESEARCH
How you'll learn
Start with guided exercises and reproduce an existing result. When you're ready, work with a mentor on a focused research question.
Seminars and reading groups
Read papers with subject experts and discuss ideas across disciplines.
Develop a literature review and project brief, then present your work for feedback.
Mentored research
Set a research question and evaluation method with a mentor. Reproduce a baseline, run controlled experiments and discuss the results.
Document your work in reproducible code, an experiment log and a research report.
AI COMPLIANCE & SAFETY
AI safety in practice
Data rights and privacy
Check consent and licenses before using data. Collect only what a project needs and protect sensitive student information.
Data-use and retention checklist
Academic integrity
Credit your sources, keep authorship clear and explain where you used AI.
Project integrity statement
Reliability and oversight
Test for bias and errors. Record the system's limits and who is responsible when it needs human review.
System card and risk register
AI security
Study prompt injection, set tool permissions and review code and synthetic security logs for problems.
Instructor-reviewed safety report
Security exercises are supervised and use approved training materials.
LLM LEARNING LAB