Principal software engineer and knowledge graph architect building the ontology and AI agent layer for Medidata's clinical trial platform (Dassault Systèmes). I design information architectures that ground LLM-based agents in structured domain knowledge — deciding how to organize data so that AI systems can reason over it reliably.
Previously lead software engineer at Diffbot, where I worked on one of the world's largest knowledge graphs (10B+ entities, 1T+ facts extracted from the open web). I hold a PhD from Leipzig University on large-scale knowledge extraction and quality assessment, co-edited the W3C SHACL specification, and co-authored Validating RDF Data (Morgan & Claypool). My research spans 50+ peer-reviewed publications with 7,500+ citations (h-index 22).
Current focus: LLM + knowledge graph integration, multi-agent AI architectures, semantic data modeling, and the evolving question of how knowledge should be structured for AI consumption.
Dimitris Kontokostas is a principal software engineer at Medidata (Dassault Systèmes), where he architects the ontology layer and AI agent infrastructure for the Medidata data fabric platform. Before Medidata, he was the lead software engineer of the Diffbot Knowledge Graph and the tech lead of the GeoPhy Data Platform. Dimitris has a PhD (Dr.-Ing.) from Leipzig University (AKSW/KILT Group) on "Large-scale knowledge extraction, publishing and quality assessment" (magna cum laude). He served as the head of technical developments of DBpedia for three years, is the creator of RDFUnit, co-editor of the W3C SHACL specification, and co-author of "Validating RDF Data."
/me @
Github,
LinkedIn,
Google Scholar,
Mastodon,
YouTube (talks),
Uni. page
or read more details on my working and academic experience in
my CV.