Aurimas Griciūnas outlines a system for developing Agentic Systems using Evaluation Driven Development, from defining problems to production integration. Key stages include prototyping, setting performance metrics, and utilizing LLMOps. The process emphasizes rapid iteration, observability, and the importance of aligning with business objectives for successful AI projects.
Table of contents
Defining The Problem.Building a Prototype.Defining Performance Metrics.Defining Evaluation Rules.Building a PoC.Instrumenting the Application.Integrating a with an Observability Platform.Evaluating Traced Data.Evolving the Application.Exposing new version of the Application.Continuous Development and Evolution of the Application.Monitoring and Alerting.Wrapping up.Sort: