Data Scientist's early models failed spectacularly in production, teaching crucial lessons about healthcare AI. Epic EHR system complexity and data leakage nearly derailed her first predictive model, which incorrectly used current lab values instead of values at prediction time. This forced her to master clinical workflows and SQL to gather proper data. Her DIVA model for difficult IV access succeeded by mimicking real-time conditions, deploying into clinical workflows and earning promotion to Data Scientist II.

Now exploring AI summarization for financial releases, her journey shows production readiness trumps notebook brilliance.