I'm starting a video series to help AI developers create applications that are reliable, accurate, and fast! It all starts with tracing your LLM application. Tracing helps you surface problems automatically by breaking down what happens when you use a large language model. You can track down issues around bad responses, prompt templates, latency, function calls, retrieval workflows, and more. I've uploaded a tutorial which helps you auto-instrument your OpenAI application, ingest traces, and use the Arize platform with a code example. Happy building!