I should also add that there is a duckdb fdw, so you could have DuckDB read from your parquet files and do faster transformations before you pull your data into Postgres!

https://github.com/alitrack/duckdb_fdw