Regina Barzilay was a leading figure in natural language processing at MIT when her own breast cancer diagnosis revealed something startling: almost none of the AI her field had built was being used to care for patients. She left her field behind to start over in medicine, working without funding or a name in the field while every early grant was rejected.

A decade later, her model Mirai can predict breast cancer years before it appears on a mammogram, and it is now part of national cancer screening guidelines. In this conversation, Regina talks with Cheraé Robinson about watching AI evolve long before ChatGPT, why bringing it into hospitals takes so much longer than building it, and the new tools she is developing for lung cancer and metastatic disease. She also speaks openly about failure and patience, and about the real story behind breakthroughs that rarely makes the headlines.