To add on this:
writing this data to parquet is not the issue here.
Just make sure the csv file is stored on a hdfs-enabled storage.
The hard part is making queries ion this parquet-data reasonably fast.  So you will probably need some performance tuning and besides bucketing/partitioning you are kinda limited in parquet.
Delta lake/Iceberg or Databricks managed tables with predictive optimization might be better choices here.
There is also parquet v2, which is not enabled by default, you might wanna look into (it has better compression).  The latter however I have not yet tested myself.