Keynote Speaker


Rajiv Krishnamurthy

Keynote Title: Data Science: 2019 and Beyond – A Practitioner’s Perspective

Time: 8:05–8:55am   Location: Grand Ballroom D

Bio: Rajiv Krishnamurthy is currently a Director of Data Science at Facebook Inc, Menlo Park, CA. He joined Facebook in 2011, since then he has led data science teams to optimize Facebook and its backend infrastructure while its userbase grew rapidly to over a billion plus users worldwide. Prior to Facebook, he worked in industry roles that focused on matching-algorithms and predictive analytics in the digital consumer space. Overall, Rajiv has been at the intersection of data, technology and decision making for the last fifteen years - first as an academic and then as a practitioner. As industry datasets evolve, tools mature and new business models emerge, he is passionate about bringing industry and academia closer to solve hard, practical problems. He has a PhD in Management Science, Masters in Computer Science, Masters in Telecom Engineering all from the University of Texas at Dallas, his undergraduate degree is in Electrical Engineering from NIT Warangal, India.


Mohsen Bayati

Keynote Title: Online Personalization of Many Decisions with High-Dimensional Covariates

Time: 2:00–3:00pm   Location: Grand Ballroom D

Abstract: A central problem in personalized decision-making is to learn decision outcomes as functions of individual-specific covariates (contexts). Current algorithms on this topic (specifically, k-armed contextual bandits with d dimensional covariates) have regret bounds that scale as polynomials of degree at least two in k and d. In this talk, we present a new algorithm (REAL-Bandit) that targets the learning towards shared similarities among decisions. We also prove that the regret of our algorithm scales by r(k+d) when r is rank of the k by d matrix of unknown parameters. REAL-Bandit relies on ideas from low-rank matrix estimation literature and a new row-enhancement subroutine that yields sharper bounds for estimating each row of the parameter matrix that may be of independent interest. Moreover, we show empirically that REAL-Bandit algorithm outperforms existing benchmarks.

Bio: Mohsen received a BS in Mathematics from Sharif University of Technology and a PhD in Electrical Engineering from Stanford University in 2007. His dissertation was on algorithms and models for large-scale networks. During the summers of 2005 and 2006 he interned at IBM Research and Microsoft Research respectively. He was a Postdoctoral Researcher with Microsoft Research from 2007 to 2009 working mainly on applications of machine learning and optimization methods in healthcare and online advertising. In particular, he helped develop a system for predicting hospital patient readmissions and obtained a decision support mechanism for allocating scarce hospital resources to post-discharge care. Their system is currently used in several hospitals across US and Europe. He was a Postdoctoral Scholar at Stanford University from 2009 to 2011 with a research focus in high-dimensional statistical learning. In 2011 he joined Stanford University as a faculty, and since 2015 he is an associate professor of Operations, Information, and Technology at Stanford University Graduate School of Business. He was awarded the INFORMS Healthcare Applications Society best paper (Pierskalla) award in 2014 and in 2016, INFORMS Applied Probability Society best paper award in 2015, and National Science Foundation CAREER award.