HBHI Workgroup on AI and Healthcare Seminar Series
Title
CliMB: An AI-enabled Partner for Clinical Predictive Modeling
Location and local date/time
This event will take place on February 21 at 17:00 GMT.
Abstract
In this talk, I will introduce CliMB, a no-code AI-enabled partner designed to empower clinician scientists to create predictive models using natural language. CliMB streamlines the medical data science pipeline, enabling users to build robust models from real-world data within a single conversation. It also generates structured reports, interpretable visuals, and automated performance evaluations, ensuring transparency and usability for clinical decision-making.
I will present findings from systematic evaluations demonstrating CliMB’s superior performance over GPT-4, particularly in planning, error prevention, and model execution. Additionally, I will discuss results from a blinded study involving 45 clinicians across specialties and career stages, where over 80% preferred CliMB for its clarity, ease of use, and reliability. By integrating advances in data-centric AI, AutoML, and interpretable ML, CliMB lowers the barrier to AI adoption in medicine, offering clinician scientists a powerful, intuitive tool to harness AI for real-world impact.
