Learning Goals

Students who complete the course should be able to:

---

Formulate Constrained AI Problems

Formulate AI problems with explicit hard constraints and identify appropriate enforcement mechanisms.

Compare Approaches

Compare Lagrangian, projection-based, and solver-in-the-loop approaches theoretically and empirically.

Critically Evaluate Safety Claims

Critically evaluate safety and alignment claims in modern ML papers.

Design Safe AI Systems

Design a constrained AI system with provable guarantees or well-justified approximations.

Contribute to Research

Contribute to the research literature on constrained and safe AI, at the level of a workshop or conference submission to ICML, NeurIPS, or ICLR.