Generative AI offers significant potential for transforming analytics, however, the foundation models, trained on readily available data, fails in situations that experts would recognize but novices would not. In analytics projects where context and nuance are important this leads to faulty analysis and failure of the project. This talk will demonstrate integrating generative AI into an analytical workflow, examining its utility across the entire project lifecycle as presented by the INFORMS Analytics Framework, while identifying some failure modes made by foundation models and proposing best practices to mitigate errors and provide better analytic insights to decision makers.
Speaker Info:
Louis Luangkesorn
Dr. Luangkesorn is a Lead Data Scientist on the Highmark Health Predictive Analytics Team where he works on projects applying statistical, predictive, operations research, and Generative AI models in use cases involving human resources and healthcare. Prior to working at Highmark Health he was an Assistant Professor of Industrial Engineering at the University of Pittsburgh. In this role he worked with industry partners including over 20+ industry based capstone projects per year for the department of Industrial Engineering and the Health Systems Engineering concentration at the University of Pittsburgh Graduate School of Public Health. He has served on the INFORMS Job Task Analysis committee, is currently on the INFORMS Practice Strategy Committee, and is President of the INFORMS Pittsburgh Chapter.