Andrew Smith is a Senior Principal Scientist in Informatics and Predictive Sciences at Bristol Myers Squibb, where he builds multi-agent LLM systems that help scientists work with complex research data. His work spans genomics, clinical, and real-world data, with a focus on making agent output verifiable enough for scientific use. He is particularly interested in where agentic systems genuinely compress research timelines, and where human expert judgment stays essential. He is a co-author of the LLM-driven CanvasXpress visualization tool published in the Journal of Open Source Software, which generates interactive charts from a conversational description rather than hand-written configuration, and he coaches BMS scientists to build AI tools of their own.