Back to projects
Production NDA — described architecturally

GraphRAG Due Diligence Tool

Knowledge-graph-powered M&A document analysis

Knowledge-graph-powered analysis platform for due diligence workflows. Automated NER, entity resolution, multi-source ETL, and multi-document summarization via LLM pipelines. Reduced analyst research cycles by ~40%.

Tech stack

Neo4jLangChainspaCyPythonFastAPIDocker

Problem

M&A analysts spending days manually cross-referencing documents across multiple sources to identify key entities, relationships, and risks.

Architecture

  • ▸**ETL:** Multi-source document ingestion (PDFs, structured data) with automated preprocessing
  • ▸**NER:** spaCy + custom fine-tuned models for domain entity extraction (companies, people, financials)
  • ▸**Knowledge Graph:** Neo4j stores entities + relationships; Cypher queries surface connections invisible in flat text
  • ▸**GraphRAG:** LangChain graph chains traverse the knowledge graph to augment LLM context with relational data
  • ▸**Summarization:** Multi-document LLM summarization with source attribution

    • Outcome

    ~40% reduction in analyst research cycle time. Surfaced entity relationships that manual review consistently missed.

    All projects