Deal Forensics Lab.

Financial analysts were losing 40+ hours per deal manually cross-referencing complex PDF Operating Memorandums with municipal records. Human error in classification was stalling major acquisitions.

I architected a forensic-grade multi-extractor pipeline using EasyOCR and custom LLM heuristics. The solution features a stateful extraction engine that maintains context across multi-hundred-page documents, streaming real-time progress via SSE to keep the UI reactive during heavy processing.