NSF awards Five Year Early Career Development Grant to Prof. Rusmevichientong

The National Science Foundation (NSF) has awarded Professor Paat Rusmevichientong a prestigious Early Career Development grant totalling $400,000 for five years. The NSF program, known as CAREER, selects those "teacher-scholars who most effectively integrate research and education within the context of the mission of their organization." About 350 such awards are granted annually to faculty throughout the United States.

Rusmevichientong, known as "Professor Paat" to his enthusiastic students, proposes research aimed at decision makers who must select, in real time, from among a large number of alternatives. The difficulty of the optimization problem for these decision makers is compounded by underlying uncertainty and by the high volume of data needed to characterize the operational situation. Rusmevichientong proposes to develop a suite of methodologies and algorithms that can be customized to exploit the special structure of each decision maker's problem. To accomplish this he will create a unifying framework for studying and analyzing this class of problems, which also encompasses the 'multi-armed bandit' problem formulated by the statistician Herbert Robbins more than 50 years ago. 

Rusmevichientong says he was motivated to develop his NSF proposal by his involvement with two companies, Analog Devices Inc. and Amazon.com.  His association with Analog Devices came through serving as advisor, in collaboration with Professor David Williamson,  for a Master of Engineering project on Google AdWords   with the Burlington Massachusetts company as the client. He says "this experience gave me an excellent exposure to the search-based advertising industry."  Working on a grant from Amazon.com in association with Professor Jack Muckstadt, Rusmevichientong obtained "real-world experience about how actual supply chains operate" and was motivated to look at data-driven real-time control of such systems. 

According to Rusmevichientong's proposal to the NSF, it is typical of problems of the sort he encountered at these companies (and that are widely found elsewhere) that there is inherent randomness in the process - for example, Amazon does not know with certainty the demand for each specific book or other product, or which products will be part of an order that must be shipped to a given customer, yet must maintain an appropriate level of inventory without tying up excessive capital. Given this inherent randomness, it would be helpful to know the probability distribution of possible outcomes, but that is typically not available.  He proposes to solve such problems in two stages, first by developing algorithms taking into account the special struture of the problem and  assuming the distribution is known and then modifying the algorithm to run in real time and adapt to new data, in effect 'learning' the underlying probability distribution over time. The complexity of the problems limits the approach to producing approximations to the optimal solution but Rusmevichientong proposes to develop mathematical bounds on how close the approximation gets to the optimal value. 

Rusmevichientong plans to use his research and experience with industry to "enrich the classroom experience for students," through experiental learning and case studies, in order to "prepare them for innovative jobs in the information technology industry." He currently is co-Principal Investigator with Professor David Shmoys on a related NSF grant, and was previously co-Principal Investigator on the aforementioned grant from Amazon.com on "Stochastic Optimization of Amazon.com's Fulfillment Network."  

A 2003 Ph.D. graduate of Stanford University, Rusmevichientong  received first prize in the George B. Dantzig Dissertation Award competition, which is named after the pioneering developer of the theory of linear programming and is awarded by the Institute for Operations Research and the Management Sciences (INFORMS). He has expressed gratitude to Professors Williamson, Muckstadt, Shmoys and other colleagues for valuable collaboration and assistance leading to the grant proposal. 

Other faculty in the Graduate Field of Operations Research who have held NSF CAREER grants include Professors Shane Henderson, Jon Kleinberg, Robert Kleinberg  and Mark Lewis (who was further selected for the Presidential Award known as PECASE) . The CAREER Award program, including PECASE, is the successor to the NSF's Presidential Young Investigator Award program that provided similar grants to Professors Robin Roundy, David Shmoys and Eva Tardos. Ten CAREER awards were granted to Cornell faculty members in 2007.