Into the Black Box: Observability in the Age of LLMs

Keynote talk exploring how to apply observability principles to LLM-powered applications in production environments. Learn how CEO and Cofounder of Honeycomb, Christine Yen, approaches the challenges of integrating Large Language Models into existing products. Discover why, despite their seemingly magical capabilities, LLMs ultimately present familiar operational challenges including performance issues, accuracy problems, and user confusion. Understand how to treat LLMs as architectural "black boxes" similar to APIs and databases, but with inherent unpredictability and probabilistic outputs. Gain practical insights on implementing observability best practices and Service Level Objectives (SLOs) in highly parameterized AI systems with nondeterministic outputs. This 15-minute CNCF keynote from KubeCon + CloudNativeCon Europe equips developers and engineers with strategies to confidently navigate the complexities of deploying and maintaining AI-driven applications in production.