Opik Tutorial | Best Practices for Evaluating AI Agent Conversations w/ Thread-Level Expert Feedback скачать в хорошем качестве

Opik Tutorial | Best Practices for Evaluating AI Agent Conversations w/ Thread-Level Expert Feedback

This is your step-by-step guide to building more auditable, outcome-aware, and human-aligned AI systems. It is time to move beyond trace-level validation and evaluate entire conversations using Comet Opik. This video walks you through how to collect meaningful subject matter expert (SME) feedback and turn it into powerful, scalable evaluation metrics, using Opik’s thread-level logging and LLM-as-a-Judge tools. Perfect for AI engineers, data scientists, and agent developers working on production-grade LLM applications. What You’ll Learn: Why trace-level evaluation is not enough How to monitor full sessions and measure real outcomes How to integrate human-in-the-loop workflows into agent development How to transform SME feedback into automated, goal-aligned metrics How to debug, inspect, and improve your agentic systems at scale Links Financial Advisor Agent Example from the Google ADK https://github.com/google/adk-samples Opik ADK integration https://www.comet.com/docs/opik/traci... Opik online thread-level evals https://www.comet.com/docs/opik/produ...