AI/MLGraphRAGBackendData Engineering
MemoryWeb / Memory-Web2
GraphRAG intelligence system combining graph retrieval, vector retrieval, and LLM reasoning for financial-crime relationship analysis.
2024
The Problem
Traditional search and RAG pipelines struggle to uncover complex multi-hop relationships across large financial filings and entity graphs.
The Solution
GraphRAG intelligence system combining graph retrieval, vector retrieval, and LLM reasoning for financial-crime relationship analysis.
My Role
AI & backend developer — designed retrieval and reasoning workflows combining graph retrieval, vector stores, and LLM reasoning.
Key Highlights
- Combined TigerGraph graph retrieval with ChromaDB/FAISS vector retrieval
- Used SEC EDGAR data as part of the intelligence pipeline
- Hackathon evaluation reported 96.4% token reduction and 0.9521 BERTScore F1 on the project benchmark
- Designed retrieval and reasoning workflow around structured relationships and semantic context
Tech Stack
GeminiTigerGraphChromaDBFAISSGraphRAGSEC EDGAR