Healthcare GraphRAG
A healthcare GraphRAG system combining semantic retrieval with structural search over a medical knowledge graph.
The problem
Healthcare retrieval needs relevant context and connected structure, not only independent semantic matches.
The approach
The project uses LlamaIndex, a PropertyGraph, PyTorch Geometric, and a VGAE with GraphSAGE to support hybrid semantic and structural scoring.
Available materials
The repository includes indexing, graph conversion, GNN training, hybrid retrieval, inference scripts, tutorials, and stored graph artifacts.
Contribution & context
Lab project on graph-based information retrieval.
Project scope & limitations
A lab implementation with offline graph training and hybrid retrieval experiments.