I have tried this before, you would need to download the file using the Amazon services module, and then create a custom Java action that uses the apache parquet library and read the data / transform to mendix objects from there.
In my case, the parquet file was converted to csv within AWS so I was able use the CSV module to load the data.