Document classification for Monetra
Automatic routing of incoming documents into 24 categories with confidence thresholds. A case study by Rahul Mehta, Machine Learning Engineer (LLMs & RAG).
Challenge
Monetra, a cross-border payments company, received 5,000 documents a week by email and upload. A team of four sorted them by hand before any work could begin.
Solution
I trained a lightweight classifier on embeddings with an LLM fallback for ambiguous cases, routed documents automatically above a confidence threshold and sent the rest to a review queue.