Knowledge graphs often suffer from low-quality, ambiguous data. In this talk, Carl will demonstrate an agentic approach to building high-quality, disambiguated knowledge graphs from messy, unstructured data. By integrating Memgraph with a custom-built agent pipeline that leverages LLMs and vector search, Carl has developed a robust system for entity resolution and graph curation. 

The presentation will showcase the practical challenges of graph construction, the need for context-aware disambiguation, and the benefits of human-in-the-loop oversight.

Carl Kugblenu is a senior software engineer at WithSecure and a PhD researcher at Aalto University, specializing in reputation systems and knowledge representation. He is passionate about knowledge graphs and AI, with a focus on building intelligent, structured systems.

From Chaos to Context: Building Knowledge Graphs with AI Agents and Vector Search