All work
LLM & NLPResearch2025
HybridRAG — vector + graph retrieval
A retrieval system that classifies the question first, then picks between semantic search, a knowledge graph, or both.
Scope something similarCode walkthrough available on request
The problem
Plain vector RAG answers 'what is X' well and fails at 'why does X relate to Y'. Scientific literature is full of the second kind of question.
What we built
- Query classifier that sorts questions into explicit facts, implicit reasoning, hidden rationale and interpretable rationale
- VectorRAG path using dense embeddings for semantic similarity retrieval
- GraphRAG path over a Neo4j knowledge graph extracted from the source papers, for relationship and multi-hop reasoning
- Context fusion that merges structured and semantic evidence before generation
- Full ingestion pipeline: PDF extraction, cleaning, chunking, embedding and graph construction, plus RAGAS-based evaluation
Need something like this?
We'll tell you honestly whether your case is a two-week build or a three-month one — and what it would cost — before you commit to anything.
