Sift

Any document in. Clean, structured data out.

Overview

Sift turns PDFs, Word documents, spreadsheets, and scanned files into clean, usable data. Upload a document and Sift parses its contents, preserves its structure, uses OCR when needed, and returns the fields or JSON you need without manual cleanup.

It handles the work between receiving a document and using the information inside it, so the output can go directly into your app, workflow, or database.

Key Features

  • Any document in: Upload PDFs, Word documents, spreadsheets, and other supported files.
  • Structured data out: Extract fields, tables, summaries, Markdown, or JSON from a single document.
  • OCR when needed: Automatically handle scanned and image-based documents without requiring a separate OCR workflow.
  • Schema-based extraction: Tell Sift what information you need and receive typed, structured fields in return.
  • Structure preserved: Keep headings, reading order, paragraphs, tables, and document relationships intact.
  • Ready to use: Get output that can move directly into downstream systems without additional parsing or cleanup.

Tech Stack

  • Backend: Python extraction pipeline with pytest coverage
  • Frontend: Action bindings for review and extraction controls

Status

Active collaborator work on reliability: clean pytest runs, frontend binding refactors, and cleanup of stale OCR engine paths while the feature set grows.