Conceptual problem with inserting csv “large data” into database and data cleaning

Question

I have a big csv files from Kaggle (200 MB and 3 GB). The data inside csv is multi type (text, floats, integers) and "dirty" (empty fields, undecoded text signs, etc.)