Harish Haresamudram

Georgia Institute of Technology

Harish’s research broadly involves learning representations for time-series sensor data, as collected from movement sensors like accelerometers and gyroscopes onboard wearable devices including smartwatches and smartphones. With a special focus on techniques that require minimal human supervision, he develops unsupervised and self-supervised methods for Human Activity Recognition (HAR) and for behavior analysis, where human movements can be used to derive insights into routines and changes in behavior patterns.