Reading through the documentation for pandas.json_normalize the record_path takes a str variable.

Thus you would need to create two dataframes for each record path and then merge them together based on a common field.