Fine-Tuning & Customisation
Fine-tuning adapts a pre-trained model to a specific task, style, or domain using your own data. It is often misapplied — many teams spend weeks fine-tuning when better prompting or RAG would have solved the problem faster and cheaper. This section explains when fine-tuning is genuinely the right tool and how to do it efficiently.
In This Section
Prompting vs RAG vs Fine-Tuning
The decision framework — when each approach is the right tool, ordered by cost and complexity. Fine-tuning is usually the last resort, not the first step.
LoRA & QLoRA Approaches
Parameter-efficient fine-tuning using Low-Rank Adaptation — how it works, why it makes fine-tuning accessible on consumer hardware, and GGUF quantisation for deployment.
Hosted Fine-Tuning Services
OpenAI fine-tuning, Google Vertex AI tuning, Together.ai, and Hugging Face AutoTrain — managed services that remove the infrastructure burden.