How to build high-quality, domain-specific training datasets for LLM fine-tuning using structured web scraping. Best practices for data quality, compliance, and scale.