Feb 3
Loading video player...
Learn about evaluating LLMs and agentic systems, with a practical end-to-end framework that shows how to combine qualitative review, structured human evaluation, and benchmarks to measure what matters in production. Large language models and agentic systems are moving quickly from prototypes into production, but knowing how to evaluate them effectively remains one of the biggest challenges teams face. In this recording, we explore the full spectrum of LLM and agent evaluation approaches, from lightweight qualitative reviews and “gut checks” to structured human evaluations and automated benchmarks. Rather than framing these methods as tradeoffs, we’ll show how they work best together across different stages of development. We’ll dig into where human judgment is essential: evaluating usefulness, reasoning quality, safety, and alignment with real user needs. You’ll learn why benchmarks alone often fall short, how to avoid common evaluation pitfalls, and how to incorporate human review at scale without slowing teams down. You’ll walk away with: • A practical framework for evaluating LLMs and agentic systems end to end • Clear guidance on when to use benchmarks vs. human evaluation • Strategies for scaling human review while maintaining rigor and speed • A better understanding of how to measure what actually matters Whether you’re building, deploying, or managing AI systems in production, this video will help you design evaluation pipelines that deliver real insight and confidence.