Lead AI Engineer, Knowledge

Select how often (in days) to receive an alert:

Lead AI Engineer, Knowledge & Semantic Systems

Bausch + Lomb (NYSE/TSX: BLCO) is a leading global eye health company dedicated to protecting and enhancing the gift of sight for millions of people around the world—from the moment of birth through every phase of life. Our mission is simple, yet powerful: helping you see better, to live better.

Our comprehensive portfolio of over 400 products is fully integrated and built to serve our customers across the full spectrum of their eye health needs throughout their lives. Our iconic brand is built on the deep trust and loyalty of our customers established over our 170-year history. We have a significant global research, development, manufacturing and commercial footprint of approximately 13,000 employees and a presence in approximately 100 countries, extending our reach to billions of potential customers across the globe. We have long been associated with many of the most significant advances in eye health, and we believe we are well positioned to continue leading the advancement of eye health in the future.

Overview

The Lead AI Engineer, Knowledge & Semantic Systems will build and maintain the foundational data, knowledge, and retrieval systems that power enterprise AI solutions. This highly technical individual contributor role focuses on creating trusted semantic models, knowledge graphs, metadata frameworks, and retrieval capabilities that enable accurate, scalable, and compliant AI-driven insights across the organization. You will help shape the data and knowledge foundation supporting enterprise AI capabilities while collaborating across data engineering, AI, analytics, and business teams to deliver innovative solutions that improve decision-making and business performance.

Key Responsibilities

• Design, build, and maintain the governed semantic layer for commercial; canonical metric definitions, dimensions, hierarchies, joins, grain, and row- and role-level access policy so that agents and BI consume one certified definition rather than reconstructing business logic per query.
• Codify commercial metric definitions including TRx, NRx, NBRx, NC TRx, market share and market basket, decile and segment assignment, and projected versus captured volume, capturing source system, calculation, refresh cadence, and known limitations for each, and arbitrate conflicting definitions across brand, field, market access, and finance.
• Build constraints that prevent semantically meaningless queries; combinations that are technically joinable but produce misleading results, such as blending projected and actual volumes, mixing incompatible grains, or computing share across non-comparable sources and enforce them in the layer rather than documenting them and hoping.
• Design and extend the commercial ontology and knowledge graph, covering ECP, account, affiliation and parent-child hierarchy, payer and plan, formulary status, product, territory and alignment, content asset, and engagement, modeling the relationships agents traverse rather than leaving them to be inferred.
• Work with enterprise team and apply entity resolution to Semantic definitions across sources used for Agents, reconciling prescriber and account identity across prescription, claims, specialty pharmacy, and CRM feeds, maintaining group practice and affiliation structures, and handling the reality that an eye care professional may be prescriber, account contact, and point of dispense simultaneously.
• Maintain taxonomies, controlled vocabularies and synonym and alias mappings so that varied business phrasings resolve onto the same governed concept, and so agents disambiguate consistently across users and brands.
• Encode temporal validity for territory realignments, formulary changes, affiliation changes, and segment refreshes, so historical questions return historically correct answers rather than today’s structure applied to last year’s data.
• Own the retrieval stack end to end; chunking strategy, embedding model selection and evaluation, index configuration and tuning.
• Treat retrieval quality as a measured discipline, building labeled query sets and tracking recall and precision at k, groundedness, and citation accuracy, with regression testing so that an embedding, chunking, or corpus change cannot silently degrade every downstream agent.
• Manage reindexing, embedding drift, model migration, and the cost and performance of vector infrastructure at production scale.
• Expose semantic, graph, and retrieval capability to agents through well-specified interfaces; Model Context Protocol (MCP) serv