Safety and Evaluation Track

This itinerary is for teams that need rigorous quality and safety controls from day one.

Sequence

1. Agent Benchmarks
2. Process-Level Evaluation
3. Agent as Evaluator
4. Consistency Metrics
5. Guardrails and Output Validation
6. Prompt Injection
7. Security Design Patterns
8. Constitutional AI
9. Emergent Misalignment

Outcome

At the end, you should be able to define evaluation gates and safety controls that are auditable and repeatable.