# MLOPs and Guardrails Day 2 Lecture

> In this lecture, Zach dives into the critical topics of evaluations and reliability in AI, emphasizing that AI agents are not magical black boxes but statistical models that need careful management. He highlights how shifts in user behavior, code updates, and data changes can break AI systems and s…

- Web page: https://www.dataexpert.io/lesson/ml-ops-day2-lecture-oct25-p737-l1524
- Program: [AIExpert](https://www.dataexpert.io/program/ai-expert)
- Module: Week 4: MLOps, Guardrails & Business Value
- Access: Requires enrollment in AIExpert
- Length: 27 min video
- Academy: DataExpert.io Academy

## About this lesson

In this lecture, Zach dives into the critical topics of evaluations and reliability in AI, emphasizing that AI agents are not magical black boxes but statistical models that need careful management. He highlights how shifts in user behavior, code updates, and data changes can break AI systems and shares strategies for monitoring these issues effectively. Zach walks through various evaluation methods such as ROUGE and cosine distance and shows how they can be used to optimize prompts with DSPY.

## Navigation

- Previous lesson: [MLOPs and Guardrails Day 1 Lab](https://www.dataexpert.io/lesson/ml-ops-day1-lab-oct25-p737-l1523.md)
- Next lesson: [MLOPs and Guardrails Day 2 Lab](https://www.dataexpert.io/lesson/ml-ops-day2-lab-oct25-p737-l1525.md)