Comment by Finnoid

I don’t know if this is viable but I wonder if you could package a small open source LLM and feed the data through it in chunks to scrub names. I’m sure it would add to the processing time and bunch other issues. But just a thought.