We seek to leverage data driven machine learning solutions in a variety of interdisciplinary applications from health care to material science. Recently, we have used deep learning techniques in a wide range of problems such as predicting neighborhood health markers from Google street view images, predicting pulmonary function test values from chest radiographs for the diagnosis of Chronic Obstructive Pulmonary Disease, deciphering connectivity patterns of neural circuitry from serial section electron microscopy images and determining the origins and process history of illicit nuclear material for nuclear forensics from scanning electron microscopy images.

We also actively aim to contribute to fundamental machine learning research on topics such as semi-supervised learning, domain adaptation and interpretable machine learning. A partial list of project descriptons can be found
here . Our vision is to overcome the barriers of scarcity of annotated data and the lack of interpretability of AI models to facilitate their ubiquitous adoption in interdisciplinary research and every day life.