[Event Invite] Can a graph catch tax fraud?

Hi ,

Tax fraud rarely shows up in a single document. It shows up in the connections between companies, people, accounts and payments spread across hundreds of separate files.

On Tuesday, November 3, Jacob Coles from Redfield will show the system his team built to make those connections visible. An LLM reads unstructured documents, extracts entities and relationships, and stores them in Memgraph as a single graph.

Investigators can then search that graph for multi-step relationships, suspicious payment patterns and conflicts of interest or ask it questions in plain language without deep graph-tech expertise. The LLMs and the database all run locally, so sensitive case data never leaves a controlled environment.

Jacob will run the full workflow live. You’ll see how to:

• Extract entities and relationships from raw documents with an LLM and build the graph in Memgraph
• Search the graph for links several hops apart, which similarity search over text chunks can’t follow
• Query the network through a chat interface, with no graph or Cypher knowledge needed

This session is engineers building GraphRAG pipelines, people designing knowledge graphs from messy documents, and teams whose data can’t leave their own infrastructure.

📅 Mark your calendar:

• Date: Tuesday, November 3rd, 2026

• Time: 9am PT | 12pm ET | 5pm GMT | 6pm CET

• Duration: 45 minutes

Stay Graph-spired! 🚀

Katarina Supe
Head of Developer Experience @ Memgraph

Memgraph Ltd., One Suffolk Way 1st Floor, Suite B, Sevenoaks, Kent, TN13 1YL, United Kingdom