Introducing s-1: Sievly’s new document reading model
All posts
Case Studies
How Quarryline turns research notes into a searchable knowledge base
Quarryline’s data team uses Parse and Extract to turn thousands of analyst notes, filings and spreadsheets into clean, cited records their AI assistant can trust.
Ines Okafor
·
Customer Engineering
sievly
Quarryline × Sievly
Quarryline’s analysts write and receive a constant stream of research: broker notes, quarterly filings, model spreadsheets and scanned meeting packs. The data team wanted an assistant that could answer questions across all of it, with a source for every answer.
The challenge
Their first pipeline used a general OCR tool and a large language model. It worked on clean PDFs and failed on everything else. Tables lost their headers, footnotes floated free of their numbers, and the assistant could not say where an answer came from.
The approach
Quarryline moved ingestion to Parse for layout-aware chunking and to Extract for the fields their analysts search by most. Every value now carries a citation back to the page and the box it came from.
The result
Analysts can open the exact source behind any answer, and the data team no longer maintains a library of per-format fixes. As the Head of Data Engineering put it, the pipeline finally reads documents the way their analysts do.
More from the blog
sievly
Introducing s-1
Announcements
Introducing s-1, our new document reading model
s-1 reads PDFs, scans and spreadsheets the way a careful analyst does, and returns structured data with a citation for every value. It is available in preview today.
sievly
Reading order
Research
Research note: reading order in complex layouts
Multi-column pages, sidebars and callouts confuse simple reading order. We share what we learned teaching s-1 to follow the page the way a person would.
sievly
Layout-first chunking
Engineering
Layout-first chunking: why we split by structure, not tokens
Fixed-size chunks cut tables in half and separate headers from their rows. Here is how Sievly decides where one idea ends and the next begins.