Abstract: Consumer wearable devices have seen remarkable progress in recent years, becoming more accurate, pervasive, and affordable. Despite these advancements, large-scale foundation models trained on wearable data are a relatively new phenomenon, unlike in many other domains. This talk will explore the technical landscape currently limiting progress in this area, and present our new work, "Scaling Wearable Foundation Models," recently accepted to ICLR.
Bio: Shyam Tailor is a senior research scientist at Google, specializing in the development of practical, scalable techniques for harnessing large foundational models in health. With a strong track record of both research and real-world impact, Shyam has contributed to several flagship features on Google Fitbit devices, including the first widely available GenAI-powered feature, as well as advanced sensing algorithms for heart rate and activity monitoring which have been widely applauded by the press. Shyam completed his PhD at the University of Cambridge in under three years, where his research, under the guidance of Professor Nic Lane, focused on creating resource-efficient machine learning algorithms. His work has earned recognition across major conferences and workshops, with publications at prestigious venues including ICLR, ICML, MLSys, ECCV, and Ubicomp/IMWUT, and he has won multiple best paper awards for his contributions.
