What is fine-tuning?
Fine-tuning is the process of further training a pre-trained AI model on your own examples so that it performs a specific task, tone or output format better. It changes how a model answers; it is not a good way to keep what the model knows up to date. Most business projects try prompt engineering and RAG first.
Definition
Fine-tuning is the process of further training a pre-trained AI model on your own examples so that it performs a specific task, tone or output format better.
How is fine-tuning done?
The model is shown many input-output examples, and its weights are updated to fit them. Providers such as OpenAI and Google offer fine-tuning for some of their models through an API. For open-weight models, parameter-efficient methods such as LoRA train only a small part of the model and keep costs down. After training, the model should be measured against the base model on a held-out test set.
When do you need fine-tuning?
When you need a consistent tone or brand voice, when output must follow a strict format every time, or when you want a smaller, cheaper model to match a larger one on a narrow classification task. For current, changing knowledge, RAG is faster and cheaper; for simple behaviour changes, prompt engineering usually suffices. The methods can be combined.
Data and compliance
Industry jargon, abbreviations or an in-house writing style, in Turkish or any other language, are reasonable grounds for fine-tuning. Personal data in training sets should be anonymised. If training data is sent to a provider, review international transfers under GDPR and KVKK, Türkiye’s data protection law, with your legal adviser. Training and running the model on your own infrastructure keeps the data in-house. For the underlying model types, see LLM.
Related terms
- RAG (Retrieval-Augmented Generation)RAG (Retrieval-Augmented Generation) is an AI architecture that lets a language model retrieve relevant passages from a chosen document collection, such as a company's own documents, and use them as context before it writes an answer.
- Prompt engineeringPrompt engineering is the practice of writing, testing and refining the instructions, examples and context given to a language model so that it produces consistent, accurate results.
- LLM (Large Language Model)An LLM (large language model) is an AI model trained on very large amounts of text that answers questions, summarises, translates and writes by predicting what text should come next.
- EmbeddingAn embedding is a fixed-length array of numbers (a vector) that represents the meaning of a word, sentence, document or image, so that similar content ends up close together.
Frequently asked questions
Should we fine-tune a model to teach it our company documents?
Usually not. The model would need retraining whenever documents change, and you could not trace which document an answer came from. For document-based question answering, RAG is the better fit.
How much data does fine-tuning need?
It depends on the task and the method. For a classification or formatting task, a few hundred good examples can produce meaningful results; a small, consistent set beats a large, messy one. Start with a small trial.
Let’s talk about your transformation
Tell us which work repeats every week. We will tell you honestly whether it is worth automating.
How we start
- A free first callWe pin down the problem and the goal together.
- A 2-week discovery sprintWe look at the data, the process and the tools, and decide where to start.
- A written proposalScope, architecture and timeline, in writing.