Benchmarking Doc parsing

For resumes in different layouts, file formats, do benchmarking using the following Libraries:
- Nanonets: qwen, hugging face, no handwritten 
- Docling: IBM, many modalities, tables, ppts etc
- Llama-OCR: pdf->Markdown, together-ai- API
- Unstrcurured: pdf, html, word..
- llama parse
- etc

Decide how to evaluate? Metrics for layout & extraction?

Prep report/paper/talk