Spaces:

---

extralit

/

public-demo

like 1

Running

Hi 🤗!

We're building Extralit (https://github.com/extralit/extralit), an open-source platform helping researchers extract structured data from scientific literature through ML-powered processing. As part of our GSoC 2025 program, we're developing an OCR extraction pipeline for scientific papers with complex tables and figures.

We need GPU resources to:

1. Train specialized table extraction models on scientific papers with zero hallucination
2. Benchmark various document extraction approaches (Marker, Docling, Vision LLMs)
3. Host a demo service for researchers to process their papers

Our software helps accelerate systematic reviews that typically take researchers 6-12 months to complete manually. The hosted demo will let researchers worldwide try our extraction capabilities before deploying locally, supporting open science and research accessibility.