Monitor, understand, and troubleshoot the behavior, performance, and reliability of AI systems—especially those based on large language models (LLMs) like GPT, Claude, or open-source variants.