From Lessons Learned to Agentic Futures: Our Journey at BMS
KGC 2025
•
14m
Umesh Bhatt
This presentation explores the evolution of knowledge graph initiatives within Bristol Myers Squibb (BMS), contrasting past large-scale projects (BE-FAIR, Project Spartan) with current, more agile approaches. I present factors contributing to the limited success of earlier "big graph" endeavors, highlighting challenges related to scale, maintenance, and the classic "chicken and egg" problem of utility versus investment.
We then introduce a new strategy focused on creating smaller, task-specific knowledge graphs – analogous to microservices – that leverage the power of Large Language Models (LLMs) and Agentic AI.
We now have a few use cases utilizing Graph + RAG methodologies, and demonstrates this paradigm shift, showing how LLMs accelerate graph construction, particularly from unstructured data sources.
The presentation concludes with an optimistic outlook on the future of knowledge graphs in pharma, emphasizing the feasibility and power of this modular, LLM- and agent-driven approach.
Up Next in KGC 2025
-
Chatbot: Empowering Medical Communica...
Ying Ding, School of Information, University of Texas at Austin, Professor
Conventional medical communication has been fundamentally broken, placing an increasing burden on healthcare professionals and significantly contributing to burnout and inefficiency. Physicians, nurses, and administrative...
-
A semantic approach empowering revers...
Pooja Arora, ThoughtWorks Inc, HCLS Advisory & Strategy lead
Ganesh Kumar Sathiyamoorthi, Thoughtworks Inc, Principal Data EngineerClinical trials face a high failure rate, with 90% of drugs failing to progress due to challenges in translating preclinical results into clinical outcomes. Sydney ...
-
The Future of Knowledge Graphs: A Pat...
Moderator
Jamie McCusker, Rensselaer Polytechnic Institute, Research DirectorPanelist
Mark Musen, Stanford Medicine Professor of Biomedical Informatics Research, Stanford University
Ted Slater, EPAM Systems, Managing Principal, Scientific Informatics Knowledge Engineering
