The last few years have taught us that without an efficient supply chain, the whole world can come to a halt in a few days.

Debdatta Sinha Roy, our next pathbreaker, Senior Research Scientist at Oracle, solves various challenges across the full spectrum of retail and supply chain in order to productize the algorithms for B2B retail customers.

Debdatta talks to  Shyam Krishnamurthy from The Interview Portal about working on a range of optimization and machine learning problems to drive critical decision making under uncertainty.

For students, the field of supply chain involves coming up with complex mathematical solutions through data driven models that make the world more accessible !

Debdatta, can you take us through your growing up years?

I did my elementary schooling in Dankuni, West Bengal, a suburb of Kolkata. My father was a production manager at the Coca-Cola plant in Dankuni. He was my only teacher during my childhood, and I can bet he was one of the best teachers a child can ever have. He was very passionate about science and mathematics. I learned way more from him about critical thinking and problem solving than at my elementary school. Under his guidance, I developed a strong interest in science-related subjects.

Since school facilities were not sufficiently good in Dankuni, my father became worried about the exposure I was getting there because he believed that I could excel in mathematics if given the right environment. Therefore, after clearing the nationwide entrance examination, I attended the Ramakrishna Mission Vidyapith (RKMV), Deoghar, Jharkhand, for my middle school and high school. It is highly selective and one of India’s most reputed Central Board of Secondary Education (CBSE) schools, both at the secondary and senior secondary levels. It is the oldest institute under Ramakrishna Mission, established in 1922. Being a fully residential school, this is where I got to spend my teenage years with the brightest students in the country, further grooming my skills.

What did you do for graduation/post graduation?

After finishing my schooling at RKMV Deoghar, I cleared the Indian Institutes of Technology Joint Entrance Examination (IIT-JEE). I joined the Indian Institute of Science Education and Research (IISER), Mohali, Punjab, to pursue a BS-MS Dual Degree. It is one of the premier research institutes in India at the undergraduate level. For the first two years at IISER Mohali, we had to study all four science areas, namely mathematics, physics, chemistry, and biology. We had to choose a specialization for the next two years, and after much thought between mathematics and physics, I decided to specialize in mathematics. The fifth year was focused on a master’s thesis project.

What made you choose such an offbeat, uncommon and unique career?

The academic environment at RKMV Deoghar encouraged students to focus and excel in the subject areas of their choices and interests. Obviously, it was nice to have a good all-round performance in different subjects but channeling our energies to improve our skills in specific domains was also much appreciated. Due to the interest inculcated by my father from an early age in science and mathematics, I focused on these areas in school. There were ample resources to learn advanced concepts and materials in specific domains beyond what was taught in the classrooms.

After I cleared the IIT-JEE, the choice was between pursuing engineering and sciences. My father wanted me to pursue a research-oriented career in sciences, and I also had a strong love for science and mathematics. Therefore, joining IISER Mohali for undergraduate studies was the natural step toward a research career.

How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path

At IISER Mohali, we were encouraged to do summer research projects based on our topics of interest. The Department of Science and Technology, Government of India, provides students of IISER Mohali with the Innovation in Science Pursuit for Inspired Research (INSPIRE) Fellowship for all five years of the BS-MS program. It helps attract top talents in research and keeps students motivated in a generally not-so-glamorous research career. At the end of my first year, I stayed back at IISER Mohali for the summer internship. I worked on improving my mathematical skills and learning how to prove theorems independently (this is considered one of the toughest skills to acquire). In my second year, I got acceptance from two very prestigious summer research fellowship programs: the Visiting Students’ Research Programme (VSRP) by the Tata Institute of Fundamental Research (TIFR), Mumbai, Maharashtra, and the Summer Research Fellowship Programme (SRFP) by the Indian Academy of Sciences (IAS), Indian National Science Academy (INSA) and The National Academy of Sciences, India (NASI). By then, I had also developed a keen interest in combinatorics and graph theory. So, through the SRFP, I spent the summer at the Department of Computer Science and Automation, Indian Institute of Science (IISc), Bangalore, Karnataka. In my third year, I wanted to explore the domain of mathematical economics and social choice theory, specifically focusing on mechanism design and game theory. So, I did my summer internship at the Economics and Planning Unit, Indian Statistical Institute (ISI), Delhi. After my fourth year, I decided to continue pursuing mathematical economics and social choice theory for my master’s thesis project.

I wanted to pursue further research in mathematics, so going for a PhD was the obvious next step. However, I wasn’t sure about the subdomain of mathematics to pursue my doctoral studies. Interestingly, all these areas of mathematics, namely combinatorics, graph theory, mechanism design, game theory, etc., fall under a field called “Operations research.” Operations research is the science of decision-making. It involves tools and techniques from mathematics, computer science, and statistics. We unconsciously use operations research in our everyday lives while making various decisions by trading off between multiple competing resources and constraints. So it turned out that without me knowing about operations research, I was developing my skills in this domain.

During my fifth year at IISER Mohali, I applied to operations research PhD programs in the US. After I graduated with the President of India Gold Medal for topping my BS-MS batch at IISER Mohali across all streams (mathematics, physics, chemistry, and biology), I joined the Robert H. Smith School of Business, University of Maryland (UMD), College Park, Maryland, as a PhD student in operations research. The Smith School PhD program is globally ranked by Financial Times as a top 10 PhD program among business schools. Also, College Park, Maryland, being just about 15 minutes away from Washington, DC, provided a great location to start my US journey. During my PhD, I received various research fellowships and awards from the Smith School and the UMD Graduate School, including the best dissertation proposal prize in operations research. My PhD dissertation is titled “Data-driven optimization and statistical modeling to improve decision making in logistics.”

Operations research solves problems from a variety of application domains. For example, the two of the most heavily used modern society tools that use operations research are Google maps and Google search. Google maps solve “routing” problems to help us reach our destinations in the least amount of time given the real-time traffic data, which is crowdsourced by other Google maps users. Google search solves “ad-revenue optimization” problems to show us relevant advertisements based on our search query, hoping that we would click on the links to visit the business web pages and Google will then get paid from them in return.

In my PhD research, I focused on various real-world routing problems. Here are a couple of examples. In one scenario, I helped utility companies build “robust” routes. Utility companies regularly collect usage data from meters using RFID technology. Reading meters is stochastic due to weather conditions, surrounding obstacles, interference, etc. Therefore, the route designs have to consider the uncertainty in the system. In another scenario, I helped local governments inspect road conditions for repair. Videos are taken using a camera mounted on the inspection vehicle. The routing problem is difficult because certain turns should be prohibited at the intersections to record those properly.

How did you get your first break?

When I was ready to graduate from UMD, I interviewed with recruiters at the INFORMS Annual Meeting, my field’s largest and most important research conference. After my PhD, I started my professional career as a research scientist at Staples, the largest business-to-business (B2B) retailer. I was part of the main data science team in the supply chain and transportation division.

Operations research is heavily used in supply chain. Here is an example of an operations research problem I conceptualized and solved at Staples. Last-mile delivery drivers are paid based on the number of hours worked, miles driven, boxes delivered, and stops made. Typically, deliveries for businesses are many boxes per stop, and a Staples driver would make few such stops in a day. People working from home due to COVID-19 significantly changed the delivery “demographic” for a B2B company like Staples. Delivery locations are now spread out, leading to longer routes and very few boxes per stop. It leads to immense pressure on Staples delivery operations to deliver to customers on time. Here I designed algorithms to identify specific delivery locations that can be outsourced to “on-demand” delivery services provided by Uber, Lyft, etc., to ease Staples drivers’ workload and deliver on time with the minimum cost impact.

What were some of the challenges you worked on? How did you address them?

One of the significant challenges for any PhD student is to explain our research concisely with good examples and use cases. It is critical during job search and interviews. Basically, it is essential to understand that the recruiters are not always interested in the granular details of what we have done in our PhD. They want to know how the skills one has acquired will help solve the problems they are interested in. Therefore, one needs to summarize their research and skills to be meaningful to the other party. I participated in various opportunities provided by the UMD Graduate School and the Smith School to interact with PhD students from different backgrounds and research areas to practice research communication.

Another major challenge is how to network effectively and efficiently. As a PhD student, I traveled to and presented my work at my field’s reputed international research conferences. I made full use of my opportunities at those conferences to network with fellow PhD students from other top schools, UMD PhD alumni, and other academic and industry researchers. In addition, my department (Department of Decision, Operations & Information Technologies) at the Smith School held weekly research seminars delivered by external speakers. The department PhD students were allowed to interact with the speakers during breakfast on the day of the seminars. It helped in getting insights into how to navigate the post-PhD job market.

Last but not least, as an international student in the US, it is very important to plan the graduation date and job search process properly. These two decisions depend on each other because of the transition complications from student visa to work visa. It is also crucial to find an established and stable employer willing to sponsor a work visa to enable us to work legally in the US after graduation. I was always careful about the various regulations and timelines and proactively acted to avoid any issues.

Where do you work now? What problems do you solve?

After working at Staples for almost two years, I moved to Oracle as a senior research scientist. I am in the retail data science R&D division, the largest enterprise resource planning (ERP) software provider to various B2B retailers (fashion, grocery, and other types).

My professional research career is in the field of retail, supply chain, transportation, logistics, and service operations. The methodologies employed in my research range from optimization and machine learning to data-driven decision making under uncertainty. My PhD research is fully aligned with what I am doing in my professional career.

At Oracle, I solve various problems across the full spectrum of retail and supply chain: demand forecasting, inventory and assortment optimization, pricing and promotion optimization, recommendation systems, etc. As new research methodologies are invented to solve these challenging problems, I always need to keep myself up-to-date with the academic research literature.

On average, I spend 50% of my time researching new data science, optimization, and machine learning solutions to emerging retail and supply chain problems. The other 50% is spent working with developers to productize the algorithms I am building for our B2B retail customers. The best part about doing industry research is to be able to make an impact in the real-world in real-time.

How does your work benefit society? 

As a professional industry researcher, I want to positively impact the broader supply chain field. After COVID-19, the term “supply chain” has become a part of the commoner’s vocabulary. People have realized that without an efficient supply chain, the whole world can come to a stop in a few days. This field involves solving complex mathematical problems. I want to make the world more accessible by making the supply chain more efficient, catering to the needs of both suppliers and customers.

Tell us an example of a specific memorable work you did that is very close to you!

Within six months of starting my first job at Staples, COVID-19 lockdowns happened globally. Back then, I was working on my first major research project in the industry. I was developing a state-of-the-art demand forecast model for Staples’ supply chain and delivery operations. The forecasts are used to optimize various other critical supply chain decisions.

Forecasting demand involves using historical data. The historical data holds all the information regarding customer behavior and their responses to various external factors the retailers set forth. Since COVID-19, customer demand patterns have changed drastically, and it keeps changing very quickly. So, the historical data loses its essence before the information can be used. For this reason, all mathematical models and analytic solutions have gone for a toss.

I had to reinvent the wheel to make demand forecasting work with a robust and reliable accuracy using much less amount of usable data than what is mathematically required. It is not discussed in any academic research literature because no one anticipated a sudden lack of usable data.

Your advice to students based on your experience?

It is very important to be truthful to yourself. Identify your strengths and weaknesses and accordingly pursue a career. It is crucial to stop comparing with others and focus on what you want to do as an individual. Also, it is essential not to settle down on a career path until that career gives you some “acceptable” level of satisfaction.

Future Plans?

I enjoy doing research and plan to stay in the supply chain field for the foreseeable future. But, future plans are always dynamic. They change based on the currently available information, our current situation, and expected future outcomes from different decisions we might take!