If you are on latest version try using docpth model …its a generative ai mode ehich can work on dynamic pdfs
Alternately…identify the table column names as key words and first try to loop through each oge and find whcih pages have those keywords and separate those pges
Then you can leverage form extractor if the table structure is same across
Else train a model to extract those tables and feed only the pges with those keywords and get the data
Cheers