Google Scholar page

Hi! I am a PhD candidate in the EECS department at UC Berkeley, working with Martin Wainwright and Kannan Ramchandran. I am working in the areas of statistical learning theory and game theory, with a focus on problems in crowdsourcing. My thesis committee members are Martin Wainwright, Kannan Ramchandran, Christos Papadimitriou, and Tom Griffiths. In the summers of 2013 and 2014, I interned at Microsoft Research, Redmond working with Denny Zhou in the machine learning group.

I have been awarded the Microsoft Research PhD Fellowship for 2014-16 (thanks, MSR!), and previously the Berkeley Fellowship for 2011-13.

Recent news
  • (Coming up) Talk at CMU ML/Stat/ECE joint seminar, Sept. 2016
  • (Coming up) Talk at MIT, Oct. 2016
  • "Stochastic transitivity" paper accepted to the IEEE Transactions on Information Theory, Sept. 2016
  • "Double or nothing" paper accepted to the Journal of Machine Learning Research, Sept. 2016
  • (Coming up) Invited talk at Asilomar, Nov. 2016
  • (Coming up) Invited talk at Allerton, Sept. 2016
  • Presented paper "Feeling the Bern" at ISIT, July 2016
  • Presented paper on "non-parametric" estimation from pairwise comparisons at ICML, June 2016
  • Presented paper mechanism design for crowdsourcing at ICML, June 2016
  • New paper on robust crowd labeling, June 2016
  • New paper on futility of parametric models in active ranking, June 2016
  • Outstanding Graduate Student Instructor award at UC Berkeley, May 2016
  • Talk at Stanford, April 2016

My research interests lie in the areas of statistical learning theory and game theory, with a focus on the application to crowdsourcing. Here is a brief description of some of my work.
Statistical inference from crowdsourced data

Data in the form of pairwise comparisons arises in many applications. Pairwise comparisons are free from several biases and are much faster and easier to make as compared to numeric scores. Given a noisy comparisons between various pairs of items, how to draw meaningful inferences from this data? While prior literature mostly focus on quite restrictive "parametric" models, in my work, I instead consider models go beyond these parametric notions and are much more flexible. I develop estimation algorithms and fundamental theoretical guarantees for these models.

Selected papers:

"Unique" mechanisms for obtaining high-quality data from crowdsourcing

A major problem with the data obtained from crowdsourcing is that it is extremely noisy. My work designs novel data collection mechanisms that lead to a collection of higher high-quality data. The proposed mechanisms are rooted in fundamental theory --- we show that our mechanisms are unique in that they the only mechanisms that can satisfy a natural 'no-free-lunch' axiom. Our mechanisms have a simple and interesting "multiplicative" form.

Selected papers:

My Google Scholar page





      At UC Berkeley:
          Statistical Learning TheoryAdvanced Topics in Learning and Decision Making
          Theoretical StatisticsAdvanced Theoretical Statistics
          Convex OptimizationRandom Processes in Systems
          Computational Biology

      At IISc:
          Detection And Estimation TheoryGraph Theory and Combinatorics
          Communication NetworksInformation Theory
          Error Control CodesAlgebra
          Wireless CommunicationWireless Networks
          Space-Time CodingDigital Communication