Assistant Professor, Computer Science Department
227 Hackerman Hall
Johns Hopkins University
3400 N. Charles St.
Baltimore, MD 21218
Phone: (410) 516 4126
E-mail: ssaria | at | cs.jhu.edu
My research interests include inference and prediction in complex, heterogeneous dynamical systems, graphical models, machine learning and health systems engineering. I am particularly interested in helping solve the trillion dollar question of how can we fix our healthcare system. I develop novel ways to capture and analyze our interactions with the health care system to help make inferences about the health of an individual as well as the health system. The overarching goal is to identify opportunities and develop tools to improve the delivery of care.
On the computational front, time series data captured from passive observation of such systems present numerous challenges. The data is often high-dimensional, heterogeneous (captured from multiple measurement modalities with varying noise properties), unstructured (notion of what the right representation is often unclear) and there is often bias in what is observed. These present numerous interesting computational challenges while attempting to solve important problems!
I'm also interested in modeling and inference challenges that arise from observational data across dynamical systems more broadly. For example, modeling user activity on a desktop (the CALO project), traffic prediction from GPS data, and activity understanding from motion-sensed data. We have multiple ongoing collaborations here. See my publication page for more examples of past work.
Jojic V., S. Saria and D. Koller (2011). "Convex envelopes of complexity controlling penalties: The case against premature envelopment." Journal of Machine Learning Research, 15: 399-406.
Saria S., A. K. Rajani, J. Gould, D. Koller and A. A. Penn (2010). "Integration of early physiological responses predicts later illness severity in preterm infants." Science Translational Medicine, 2(48).