Knowledge Graphs in the era of Gen AI

In this event, we will explore three different projects showcasing the good, the bad, and the ugly of applying knowledge graphs (KGs) to real world problems.

First up, we have CompanyKG - the culmination of the research effort at EQT Motherbrain, where the team built a large-scale heterogeneous graph - developed for fine-grained company similarity quantification and relationship prediction. CompanyKG comprises 1.17 million companies represented as graph nodes, enriched with company description embeddings, and 51.06 million weighted edges denoting 15 distinct inter-company relations. CompanyKG was successfully accepted at this year's KDD'24 conference in Barcelona.

The second presentation will focus on the collaboration effort between Nobel Prize Outreach (NPO) and EQT Motherbrain. Leveraging Nobel Prize laureate’s biographies and Nobel Prize lectures, the project used OpenAI's GPT-4 to extract meaningful keywords, and detect and merge synonyms using a simple but novel methodology, before finally merging the resulting keywords with a subgraph extracted from Wikidata.

The final presentation will focus on the work conducted at EQT Motherbrain to generate a KG for private equity (PE) using a combination of GPT-4 and Neo4j. The resulting KG was trialled across a number of different tasks, including finding similar PE deals.

Lele Cao is a Principal AI Research Scientist at King with a Ph.D. in Robotics and AI from Tsinghua University. He has 15+ years of experience at organizations like EQT, Microsoft, Alibaba, and Ericsson, and has published over 30 papers and patents in top conferences and journals. Lele is also an active mentor and reviewer for leading academic events and journals.

Andrew McCornack is a Senior Machine Learning Engineer at EQT (Motherbrain) in Stockholm, Sweden, with previous roles as a Senior Data Scientist at Spotify and data roles at Amazon. He holds a degree in Psychology with a minor in Statistics from the University of Washington.

Hans Mehlin is the Chief Infrastructure and Data Officer at Nobel Prize Outreach, where he has worked since 1994. With a background in cell biology and molecular genetics, Hans focuses on Nobel datasets, integrations, and applications.

Valentin Buchner is an ML Engineer at Anotherblock in Stockholm, with an MSc in Artificial Intelligence from VU Amsterdam. He has worked at Human Insight, Landscape AI, and EQT (Motherbrain) and has several notable publications in venues like KDD and NeurIPS.

Host or Publisher Stockholm AI