Building Agentic AI: Workflows, Fine-Tuning, Optimization, and Deployment

8

Fine-Tuning AI for Calibrated Performance

Introduction

Fine-tuning is the process of taking a powerful, general-purpose language model and updating the model’s parameter values based on examples that matter for a particular application. With this approach, we aren’t rebuilding everything from scratch, but rather teaching the model to speak a new domain’s language, follow a style guide, and learn new information while minimizing losing any prior knowledge.

Off-the-shelf LLMs tend to be great generalists, but they can still stumble on edge cases. For instance, if you ask a vanilla model to generate complex SQL queries against a proprietary schema, it might hallucinate columns or misplace joins. In previous chapters, we solved this problem by ...