GraphRAG Knowledge System
Hybrid vector/graph RAG desktop app for semantic knowledge extraction
Co-built a hybrid vector/graph RAG desktop application (team of 4) that ingests notes in Markdown, PDF, and Notion formats, extracts and canonicalizes concepts with an LLM, and renders them as an interactive 3D force-directed knowledge graph inside a brain mesh. Stabilized the FastAPI backend with shared Kuzu engine, per-request connections, and async endpoints. Built the ingestion pipeline with PDF/zip upload, duplicate detection, Notion importer, and a Milkdown WYSIWYG editor.