If you are a CTO, Chief of AI, VP of Product/Data or AI, Staff+ AI Engineers, Data Scientist or anyone tasked with getting real results from RAG beyond vector search, this is for you.

Most enterprise knowledge lives in PDFs, DOCX, and TXT and not in a graph. Unstructured2Graph changes that. In this live, technical session we’ll (re)introduce the Memgraph AI Toolkit, explain the real bottlenecks behind building Knowledge Graphs for GraphRAG from unstructured data and showcase it live.

You’ll see how Unstructured2Graph ingests documents, extracts entities (via LightRAG integration), links them with page-level provenance, and writes a connected, queryable graph into Memgraph ready for retrieval.

Why GraphRAG needs connected structure (not just embeddings) for explainability and multi-hop answers

How Unstructured2Graph turns PDFs/DOCX/TXT into nodes, relationships, and attributes with preserved provenance

Where LightRAG fits for LLM-based entity extraction

How to query the resulting graph in Memgraph

Practical guardrails: handling messy layouts, duplicate entities, ambiguous mentions, and update cycles

Turn Documents into Graph Context: Unstructured2Graph + Memgraph AI Toolkit