<aside> 💡

Use OCR to extract data from complex documents

</aside>

Building RAG with complex documents is a nightmare.

Large Language Models don't work well when documents contain tables, diagrams, and forms. Anyone who's tried knows that hallucinations are horrible, and the tools out there don't solve the problem.

Now, there’s a way to make it work with really impressive results:

Instead of directly relying on an LLM, we can have a two-step process that uses Computer Vision to break down documents and feed them to the LLM… → Tweet