Evaluation & Testing for AI Systems

How do you evaluate an LLM pipeline?

What is LLM-as-judge and what are its limitations?

What is RAGAS and what does it measure?

How do you regression test a prompt change?

How do you run A/B tests on LLM outputs?