I completed my Bachelor of Engineering in Computer Science from R.V. College of Engineering under Visveswaraiah Technological University. I then pursued my Master of Engineering in Systems Science and Automation in the Electrical Engineering Department of Indian Institute of Science, Bangalore. I worked as an Engineer at IBM Research, Bangalore. Prior to that, I spent a year at Microsoft AdCenter as a part of Click Prediction Team.

Contact Details

Varun Ravikumar Embar
115 Apt A, Leonard Street,
Santa Cruz, CA, 95060
Email: vembar@ucsc.edu

Education

University of California, Santa Cruz

PhD Computer Science 2016-present

Advisor
Lise Getoor

Indian Institute of Science, Bangalore

Master of Engineering in Systems Science and Automation 2010-2012

Grades
6.9/8.0
Advisor
Indrajit Bhattacharya
Thesis
Aligned Matrix Factorization for Multitask Clustering [pdf]
Relevant Courses
Linear Algebra and Applications, Stochastic Models and Applications, Linear and Nonlinear Optimization, Pattern Recognition and Neural Networks, Game Theory, Probabilistic Graphical Models, Topics in Pattern Recognition
Award
GATE Scholarship (2010-2012)

R.V College of Engineering, Bangalore

Bachelor of Engineering in Computer Science 2006-2010

Grades
82.3% (First Class with Distinction)
Thesis
Sihi - Kannada based Programming Language
Relevant Courses
Data Structures, Analysis and Design of Algorithms, Graph Theory & Combinatorics, Operations Research, Discrete Mathematical Structures, System Modelling and Simulation
Award
Charles E Sprock Scholarship (2006-2010)

Work

IBM Research, Bangalore

Estimating Properties from Network Diffusions
I proposed a bayesian framework for evaluating network and cascade properties in the context of diffusion processes. It characterizes properties based on how efficiently they can be computed. Monte Carlo approximation and Gibbs sampling is used when properties cannot be computed efficiently. A Map-Reduce version of the algorithm allows it to handle large data. (ACM SIGKDD 2014)


Influence Analysis during 2014 Wimbledon Tennis Championship
I built a system to compute influence scores based on Twitter feeds during the 2014 Wimbledon championship. The scores were computed for two time windows in near real time; for the past 24 hours and since the beginning of the tournament. These Scores were used by Wimbledon Social Command Center present at the venue to give All England Lawn Tennis and Croquet Club(AELTC) insight into social media. This project was reported in online media like BBC and Yahoo News. (ACM SIGKDD 2015)


Topic Modeling for Exxon Mobil
I developed a module to run Latent Dirichlet Allocation (LDA) on SPSS Modeler 16 using R programming language. After basic preprocessing is done in SPSS, LDA is run in an R node and makes use of the topicmodels package. The output of LDA can be used by the SPSS stream for further analysis.


Marketing Offer Design Analytics
IBM Interact allows personalized o.ffers to be delivered to the customer when interacting with a website. Theoffers were modeled as a set of attributes such as font size, colour etc. Learning the Click Through Rates of o.ffers at the attribute level helps in providing insights to the designer about which set of attributes provide better results. (RTE Top Poster Award)

Microsoft

Click Prediction, Bing Ads
I built models to predict probability of click given an ad, user and query. To find the right set of features large amounts of data had to be queried on a hadoop-like system using SCOPE language.

Yahoo Labs!

Frequent Pattern Mining
I worked on the problem of classify spam comments using Frequent Pattern Mining. I developed an algorithm to perform frequent pattern mining using Apache Pig.

Publications

Scalable Structure Learning for Probabilistic Soft Logic
Varun R Embar*, Dhanya Sridhar, Golnoosh Farnadi, Lise Getoor

2018 IJCAI Workshop on Statistical Relational AI (StarAI) [pdf]

Aligning Product Categories using Anchor Products
Varun R Embar, Golnoosh Farnadi, Jay Pujara, Lise Getoor

2018 WSDM Workshop on Knowledge Base Construction, Reasoning and Mining (KBCOM) [pdf]

Online Topic-based Social Influence Analysis for the Wimbledon Championships
Varun R Embar, Indrajit Bhattacharya, Vinayaka Pandit, Roman Vaculin

2015 International conference on Knowledge discovery and data mining (SIGKDD)[pdf]

A Bayesian Framework for Estimating Properties of Network Diffusions
Varun R Embar*, Rama Kumar Pasumarthi*, Indrajit Bhattacharya

2014 International conference on Knowledge discovery and data mining (SIGKDD) [pdf]

Pattern clustering using cooperative game theory
Swapnil Dhamal, Satyanath Bhat, KR Anoop, Varun R Embar

Centenary Conference, 2011, Indian Institute of Science, Bangalore [pdf]