We're hiring!

- Nonparametric Discovery of Human RoutinesDesigned and implemented nonparametric Bayesian framework for human activity/routine discovery using multimodal sensor data. Our approach automatically discovers latent human r!outines without the need of predefined model parameters.-Topic Modeling for Group-buying Data AnalysisAdopted topic models with modifications to learn users’ preference and take into account u!sers’ followship, co-buying records to provide better purchase prediction accuracy.-User Demographic Profile PredictionApplied semi-supervised learning to the problem of predicting demographic profiles from browser interaction history (e.g., scrolling or typing in address bar) and achieved 72% gender prediction accuracy on a dataset of 7,600 users.

Proposed an approach that encodes the local appearance descriptors transformation byusing visual experience as a graph for the traditional feature matching scheme for large scale image retrieval in a wide viewpoint setting.

Prototyped a mobile system to monitor users’ physiological measurements and activity level.Applied pattern recognition methods to monitored signals to interpret users’ mental state. Filed a U.S. patent and published two conference papers.

Developed an end-to-end medication tracking solution from IEM (Ingestible Event Marker) tobackend server. Designed and prototyped on-body sensor and mobile application for body sensor network.