OCR and Categorization with LLMs

I'm looking for someone to write a program in python that will accept several types of documents from us in different formats that will read it, pull off of the document what we need, then store it in a spreadsheet and database. I also need it to have error logic, i would like it to attempt every document 3x then only store the data if 2 or all of the outputs match. If not it should try again in different sessions up to 3x times then push them out out to an error log. There are specific pieces of data that we need from the document. I need someone that can meet with us on zoom to work through problems, I need someone that can test well to ensure that these are done properly, and I need someone who can support this after it is done. I also need the configuration to be easily updated if new models come out that we would like to upgrade to. It should work with OpenAI, llama, and mistrial at the very least. If you have read this put LLMs at the beginning of your response.