During a data transformation process, you need to split a large dataset into smaller, more manageable subsets based on specific criteria. What techniques would you use to efficiently split the data while ensuring that the integrity and structure of the original dataset are preserved?