Automated Information Extraction Pipeline for Documents 

KritiKal Solutions developed an automated solution that efficiently manages large volumes of documents. It is powered by 8B parameter open-source Large Language Models such as OpenChat, Starling, Phi-3, Mistral and Intelligent Document Processing. The solution was trained on partly original dataset provided by the postal service provider as well as synthetic documents developed using Generative AI to identify vast types of key-value pairs, etc. When implemented, the system was able to understand, extract and process information present in invoices. It provided swift classification, validation of details, present in invoices, searching and querying operations etc. The solution was able to reduce manual processing time and human errors in invoice handling faced by the client.