Crafting the Future with Synthetic Data

On November 7, CAAI hosted Dr. Ryan Kappedal, ’19, a Booth alumnus and Technical Lead Manager at Google, for an insightful discussion on the evolving landscape of AI and the critical role of data quality and synthetic data. Dr. Kappedal’s career trajectory is nothing short of remarkable. A retired Air Force Lieutenant Colonel, his service spanned roles as an Intelligence Officer, Professor of Statistics, and Chief Data Scientist in Iraq. He earned a PhD in Statistics from the University of Washington, where he honed his expertise in data analysis and modeling. Following his transition out of the military, he completed his MBA at Booth, then navigated industry roles at Gusto and Auris Surgical Robotics before joining Google.

Dr. Kappedal’s expertise centers on synthetic data—an innovative solution to the persistent challenge of data quality in AI. As large language models (LLMs) like Google’s Gemini demand trillions of high-quality data tokens, synthetic data has emerged as a game-changer. Dr. Kappedal illustrated how synthetic data enables privacy-compliant, scalable, and statistically robust training datasets by employing methods such as differential privacy and synthetic user data generation. He shared a glimpse into how this data is evaluated using calibrated rubrics and annotations, scaling from human raters to automated processes.

Dr. Kappedal also offered a compelling vision of the immediate future in AI, emphasizing that while annotation and rephrasing are largely solved problems, areas like agentic flows and API integration remain ripe for innovation. His advice to the audience was both pragmatic and inspiring: focus on first principles, embrace evidence-based science, and remember that persistence often trumps talent. He concluded his talk with, “every good story begins with an outlier,” and his own journey is a testament to that belief.