About
I
am (edit 2013: I've finished the Ph.D.) a Ph.D. student in the
Machine Learning Group within the
Department of Computer Science at the
University of Toronto, where I also received my M.Sc.
in January of
2008.
I graduated from
Stanford
in June of 2006 with a BS in
Computer Science.
I was on leave, working as a machine learning consultant, from August 2008 - August 2009.
My advisor is
Richard Zemel;
I also work frequently with
Ryan Adams and
Brendan Frey.
In January 2013, I will be joining
Microsoft Research Cambridge as a Postdoctoral Researcher
and will be a Research Fellow at
Darwin College at the
University of Cambridge.
Broadly, I'm interested in applying machine learning to problems where examples come in the form of complex, structured
objects. How can we best learn by example when our examples are things like image segmentations, bipartite
matchings, paths in a graph, or (perhaps someday) computer programs?
What computational tools and models do we need to be able to tractably work in these settings? How can
we better model these settings without sacrificing computational efficiency?
Some specific interests:
- Inference in discrete graphical models: message passing, network flows, and linear programming;
message scheduling and dynamic inference.
- Structured output learning; complex loss functions on structured output spaces.
- High(er) order models.
- Computer vision, specifically with respect to image-labeling models.
Some selected past projects and positions:
- Research Intern at Microsoft Research Cambridge (UK) (2010)
Supervisor: Tom Minka.
improved message passing schedules for chain-like and grid-like models in
Infer.NET.
- Consultant at The Walt Disney Company, Research & Development (2009)
organized data and built
models of energy consumption in Disney theme parks.
- Consultant at Risk Management Solutions, Research & Development (2009)
modelled business interruption
losses due to hurricanes hitting oil pipelines in the Gulf of Mexico.
- Analytics team volunteer at Obama for America 2008 Presidential Campaign (2008)
analyzed and optimized www.barackobama.com and store.barackobama.com.
- Software Engineering Intern at Google (2006)
developed a new ranking method for the iGoogle gadget directory.