jargon

Applied AI·Fine-tuning and customisation

you spend three weeks collecting, cleaning and deduplicating six hundred examples, and one afternoon actually running the training.

Dataset curation

Draft summary, pending review

The real work of fine-tuning: collecting, cleaning, deduplicating and quality-filtering examples. A few hundred excellent examples reliably beat tens of thousands of scraped mediocre ones. Expect this to consume most of the project's time.