How do you handle PDF tables and scanned images in a RAG pipeline?

Tanishka's answer in Gaurav Sen's mock interview is to change only the ingestion pipeline. Common parsers such as PyPDF and PyMuPDF handle tables poorly, so a layout detection step first labels each page region as a title, table, text or logo. Each element type is then sent to a parser built for it, since one parser rarely extracts everything well.

• The interviewer's examples include currency conversion tables, policy rules and screenshots of tables pasted into PDFs.
• Tanishka notes tables can be hierarchical, with several rows nested under a single row.

▸In their words

“a single parser might not be able to you know extract all of these things very efficiently” Tanishka · at 6:35 — open in the transcript →

Source: our own transcription of the episode. The full conversation is on Gaurav Sen (2026-09-22). Read the whole episode summary →

▸More from this episode