Abstract: In this talk, Dr. Isabel Fulcher recounts her transition from academia to building an AI-powered pregnancy care platform to improve maternal health outcomes. As a biostatistics graduate student, she specialized in building perfect models: robust to misspecification, free from measurement error, and asymptotically efficient. But a perfect model helps no one if it never leaves a Jupyter notebook. At Delfina Care, building the model was only one part of the goal: train it at Mayo Clinic, validate it in Texas, deploy it live in Epic, and embed it in clinicians' workflows. Dr. Fulcher details the work required before and after model development. First, she describes the technical infrastructure needed to bridge fragmented health data systems and bring predictions into real clinical settings. Second, she draws on principles from implementation science to deliver model results to clinicians in ways that are explainable, transparent, and actionable. Third, she highlights the importance of rigorously evaluating each step of deployment: external validation, model monitoring, clinical uptake, and, ultimately, patient outcomes. Drawing on examples from reproductive health, the talk will resonate with researchers hoping to bring clinical AI out of the notebook and will highlight the central role biostatisticians can play in making it work.
About the Speaker: Isabel Fulcher, PhD is a cofounder and Chief Scientific Officer at Delfina Care, where she leads the full lifecycle of clinical AI, from model development to live deployment to prospective evaluation. Her work spans technical infrastructure for health data integration, implementation research with clinicians, and quasi-experimental methods for evaluating impact on patient outcomes. Her team's deployment pipeline supports more than 30,000 patients, and its provider dashboard is used by over 500 clinicians. Dr. Fulcher has been recognized as a STAT Wunderkind and a Top 50 Woman Leader in Healthcare Technology, and she received the American Statistical Association's Annie T. Randall Innovator Award. She was a Harvard Data Science Initiative postdoctoral fellow and has a PhD in Biostatistics from Harvard University, where she developed novel statistical methods for causal inference in complex data settings.