Research Interests

I am generally interested in theoretical machine learning, especially statistical and computational issues relevant to the use of machine learning in practice. In my time at Microsoft Research and at IIT Bombay, I have had the chance to explore both practical and theoretical problems in machine learning:

• Provable Non-convex Methods for Robust Learning
  In these projects, I worked on two problems in robust estimation. We developed non-convex optimization techniques for both these problems with similar guarantees to best convex optimization based methods with vastly superior empirical performance on practical problems.

• Entity Linking and Disambiguation
  In this project, I worked on the problem of tackling No Attachment (NA) mentions in the Entity Linking task using hierarchical non-parametric topic models.