Umesh Bhatt is Director of AI/ML & Data Engineering at Bristol-Myers Squibb, where he has spent 14 years turning knowledge management from a filing problem into a reasoning problem.

He built BMS's Semantic Data Engineering function from scratch — knowledge graph infrastructure that connects proteomics, drug repurposing, and structure-activity data across the research portfolio. More recently, he created BMS Discovery, a governed agentic research workspace that gives computational scientists natural-language access to curated scientific data services, producing versioned, auditable research dossiers. The system doesn't just retrieve — it reasons, traces evidence, and builds institutional memory that survives personnel changes and program pivots.

His published work examines why AI systems that work in demos fail in organizations. The Trust Gap — the distance between a working system and a used one — is, in his view, the central unsolved problem in enterprise AI. Not the models. Not the infrastructure. The human and organizational factors that determine whether knowledge actually flows.

He also created Epistract, an open-source biomedical knowledge graph extraction system (published on bioRxiv) that replaces similarity-based assembly with comprehension-based extraction — typed, directional, evidence-backed relationships from scientific literature, grounded in 40+ biomedical ontologies. Umesh represents BMS at the Pistoia Alliance, Allotrope Foundation, Knowledge Graph Conference, and AWS re:Invent. He writes at medium.com/@8thcross.