MS&E 318: Constrained and Safe AI

Advanced PhD seminar on algorithmic methods for safe and constrained AI

View Schedule View on GitHub


Course Overview

How can we design AI systems that are not only powerful but also provably safe and trustworthy? This advanced PhD seminar surveys algorithmic methods to enforce hard constraints in machine learning, reinforcement learning, and generative AI.

Topics include:

  • Classical constrained optimization (Lagrangian methods, robust and stochastic programming)
  • Safe reinforcement learning (trust regions, Lyapunov functions, reachability)
  • Hybrid ML-optimization methods (projection networks, solver-in-the-loop architectures)
  • Alignment strategies for large language models (fine-tuning, model editing, tool use, and interactive alignment)

We will consider applications to robotics, finance, healthcare, energy, and personal AI assistants. Topics may change according to student interest.

Prerequisites: Optimization at the level of CME 307 or EE 364a.

Course Format

For each topic, we will begin with a foundational mini-lecture, followed by student-led discussion on modern research papers. The course will culminate in a final research project which will constitute a majority of the grade.

Quick Information

  • When: Tuesdays, 1:30-4:20pm, January 9 - March 10, 2026
  • Where: McCullough 122 (or Zoom by arrangement)
  • Instructor: Prof. Udell
  • Office Hours: Schedule here

Upcoming deadlines

  • Sign up for a topic on our schedule by writing your name next to a paper 11:59pm Friday 1/9. It’s ok if >1 person signs up for the same paper, though we may have to swap some signups to make sure every paper is covered.
  • Class next week will be a working meeting: in your topic group (everyone signed up for the same week’s presentation), you’ll collaborate to pick 5-10 papers on your topic, with 2 to emphasize, and develop a baseline presentation (due Friday 1/16).

MS&E 318: Constrained and Safe AI | Stanford University | Winter 2026

This site uses Just the Docs, a documentation theme for Jekyll.