I give a small number of talks each year to senior leaders and organizations that share my commitment to improving health care. These talks examine AI’s huge opportunities and new dangers, with a focus on the decisions leaders face: where to invest, what to build, and how to tell whether it works. I draw on my clinical practice, my research at Berkeley designing and deploying AI for life and death decisions, and my experience cofounding a venture backed startup (Dandelion). Alongside building AI, I study its failures: my research on algorithmic bias has shaped health care practice and public policy. TIME Magazine recognized my work by including me among the 100 most influential people in AI.
How access to data can improve the accuracy and fairness of medical AI.
How clinical problems led me to AI, why data access limits what companies can build, and how buyers can assess whether their products work.
| Transforming health systems with AI | AI can improve individual decisions and change how care is organized. Using cases in drug treatment and heart attack testing, I show how data can guide care and payment. I then examine what happens when we organize care around those capabilities: collecting information first, then directing clinical attention where it is needed. |
|---|---|
| Bedside to bench: Reinventing medicine with AI | Medical images and waveforms contain information that researchers have barely begun to use. I explore how algorithms can find patterns in these data that doctors miss, opening paths to biological discovery and drug development. These discoveries create opportunities for companies building medical tools, but depend on access to data for development and evaluation. |
| Dissecting algorithmic bias | An algorithm can predict costs accurately and still overlook patients who need care. Drawing on my research into racial bias, I explain why predicting spending can produce this error and how choosing a better measure of patient need can correct it. This case gives leaders a way to assess whether an algorithm is solving the problem they intended. |
For speaking inquiries, contact me.