Hey HN, I'm Raunak from Reducto (https://reducto.ai), a high-quality document ingestion API tailored for language models. We developed Reducto to address our own need - no existing parsing solutions provided the accuracy and speed necessary for our user-facing AI applications. We designed a system that comprehends documents visually (like a human), ignoring document metadata and processing each page as an image to ensure the highest possible accuracy (with benchmarks to prove it).

Please give our demo a try with some of your own PDFs or reach out at founders@reducto.ai if you’d like to start using Reducto in production.