Built an AI-powered drug information retrieval system called Monographer for Merck. Healthcare professionals ask questions in natural language, system finds relevant information from vast medical databases.
Drug information is scattered across monographs, clinical studies, interaction databases, regulatory documents. When a doctor needs to know if Drug A interacts with Drug B in patients with Condition C, they shouldn’t have to search five different systems.
The system uses transformer models and semantic search to understand medical queries. Not just keyword matching—it understands that “contraindications for elderly patients” and “should I avoid prescribing to seniors” mean the same thing. Results are sourced, ranked by relevance, presented with the context medical professionals need.
Medical AI has zero tolerance for hallucination. When a doctor asks about drug interactions, a wrong answer could kill someone. We built extensive validation and source-linking to ensure every answer was traceable to authoritative sources. No made-up answers, period.