🔧 What is 'fine-tuning'?

Fine-tuning is taking an already-trained model and teaching it something extra, using a small amount of your own data. Think of it as taking a graduate and putting them through a short, specialised training course for *your* company.

Training a model from scratch is like raising a child from birth — you need huge amounts of data and money. Fine-tuning is like the two-week onboarding: the model already knows language and general knowledge, and you show it your specific examples so it learns your tone, your domain, your format.

A real example: you have a support team and want AI to answer tickets in your brand voice. You'd fine-tune a model on a few thousand of your best past answers, and it learns to write like your team.

You probably don't need to fine-tune anything today. For most people, the modern alternative is easier and better: write a good prompt with your style guide, or use 'retrieval' — feeding your documents to the model as context. Fine-tuning is for when you need the model to *consistently* behave a certain way, at scale.