[bluedemon]

Perhaps look into using dlt from https://dlthub.com, using pyarrow or polars. It handles large datasets well, especially when using generators to process the data in chunks.