Background

My interest in research started during my Masters at IIIT Hyderabad (around 2010). I did an internship at Carnegie Mellon University, and that helped me get an idea on what good research would entail. After my masters, I worked as a research engineer at Yahoo Research and in QCRI for almost 3 years, where I had the opportunity to work with amazing people and great data. Below, I try to summarize the areas I've worked on, during my PhD.

Graph Algorithms

One of the first problems we worked on during my PhD is grpah sparsification with constraints. We proposed a special, practical version of graph sparsification, given a network usage log (which is more prevelant in real world), proved that it is NP-hard and developed scalable algorithms to solve it (published in PAKDD 2016, and invited to a journal).

Distributed Computing

I worked extensively with Pig (and the Hadoop stack, in general) during my days at Yahoo. During the second year of my PhD, thanks to a Yahoo grant, I spent a lot of time trying to build algorithms for large scale graph problems. One such algorithm we proposed was implemented in Apache Giraph and used to solve the classic Facility Location problem. We tested our algorithm on large graphs with hundreds of millions of nodes (Tumblr social graph) on the Yahoo cluster (published in CIKM 2015).

Polarization on Social Media

After almost two years exploring various topics, I zeroed in on my PhD thesis topic, which deals with (i) automatically identifying polarized discussions on social media, (ii) quantifying their severity and (iii) building algorithms to reduce the polarization. This topic spans several fields, including social science, political science and psychology. We approach is from a novel, computer science perspective. It neatly ties into the recent concerns about how social media might have enabled the creation of echo chambers and filter bubbles online, since we provide an algorithmic solution to this problem. Our algorithms are mostly graph based, a deliberate choice, to avoid the use of language, which is tricky to handle, on social media. This allows our methods to be applicable for any language and any domain, unlike previous work, restricted mostly to politics.

Data Analysis Projects

One of my hobbies is to find new data sets to get insights from :). I have done a bunch of side projects during my PhD, which involved collecting lots of data, applying existing techniques, and generating meaningful insights. Examples include, invading teenagers' Twitter profiles to see how they behave when they breakup with other teens (SocInfo'14), with celebs (WebSci'17), understanding if media attention to science has increased over the last decade (WWW'17 poster), etc.

Research in practice

One of the common themes of my research is that I like to do is to put what ever we do into practice. I try hard to make sure that we build a demo showcasing our research, when ever possible, and this has been a repeating theme through out my career, from my masters days to the most recent WWW demo.