I am familiar traditional JSON to relational table ETL pipelines, but how does ETL pipelines differ in data lake/house when processing data dumps, such as PDF/i

I am familiar traditional JSON to relational table ETL pipelines, but how does ETL pipelines differ in data lake/house when processing data dumps, such as PDF/images?