Sanjmeet Abrol, who graduated a month ago from the Data Science Institute’s master’s program, already has a great job: Next week she’ll begin working as a Strategy Consultant for IBM’s Chief Analytics Office. The office works with IBM’s sales, finance and operations teams to solve problems relating to coverage optimization, fraud analytics, and portfolio optimization. Abrol will use a combination of data science and management-consulting skills to contend strategically with the above challenges.

She’s looking forward to beginning her career and is grateful to DSI for the myriad opportunities it provided her. While a student, for instance, she worked on a smart-cities research project that resulted in the publication of a scientific paper in a major journal – a paper for which she is the lead author.  DSI also gave her the guidance she needed to get an interview and consequently a job at IBM.

Here, Abrol talks about her studies at DSI, her background in India and the research she conducted in smart cities.

**********

Did having a master’s from DSI help you get the job?

Yes of course! At DSI, the academic curriculum, research opportunities, seminars, hackathons, etc., allow students to keep abreast of the latest developments in both academia and industry. Consequently, during job interviews, I was able to keep my best foot forward while being interviewed by lead data scientists and vice presidents from major organizations. The constant guidance and support from Rachel Cohen, Assistant Director of Student Services & Career Development at DSI, played a huge role in my success. Regarding IBM, Rachel introduced me to a senior team member at the Chief Analytics Office, who had previously been an adjunct professor at Columbia. A discussion over the phone with him turned into a first interview, which later turned into a great job offer.  

You graduated from Guru Gobind Singh Indraprastha University in India and then worked there as a Solutions Architect for Landis+Gyr (Toshiba Group). How did that work prompt you to apply to DSI?

At Landis+Gyr, I was the first person given the opportunity to be a part of the Solutions Architecture team right after graduating from college. Before me, every person they hired needed at least 10-years’ experience to join the team.

On the one hand, this meant that I faced more responsibilities and a steeper learning curve. But on the other hand, I had an opportunity to work on challenging smart-grid solutions. Modern-day meters for electricity, water and gas are similar to mobile phones. They are network-connected devices that provide energy-consumption information at high resolution and can provide remote control to other smart devices in your home or office. As a solutions architect, I was responsible for defining the functionality of communication between the smart meters and the meter-data management systems. I gained considerable expertise in data-communication protocols and data storage but overall I had limited data-analysis skills. Nevertheless I did have ideas for using the data collected within the smart grids and I had exciting discussions with my team about data science. It was this work, and this challenge, that inspired me to apply for a master’s in data science at DSI.  

You’ve said you were drawn to DSI because of its research centers. How so?

When I was looking into the master’s program I read that DSI, along with offering amazing classes, had research centers for data science. I looked at the research work being done at one of the centers – Smart Cities –  and knew that DSI was the right place for me since that is one of my main interests.   

What were some highlights of your time at DSI?

In my statement of purpose essay to DSI, I mentioned my interest in smart cities, and noted that “I have been fascinated by Professor Patricia J Culligan’s extensive work in sustainable urbanization, and I shall be honored to assist her in any capacity I can.”

And to my immense delight, my humble request became a reality: I began my master’s in September 2016 and have assisted Prof. Culligan since October 2016. Working under her direction, with Ali Mehmani (a postdoctoral research assistant at DSI) as my research mentor, I had the opportunity to work on three projects, each of which aims to optimize and reduce energy consumption in non-intrusive ways by analyzing data generated in smart buildings.

Can you discuss one of those projects?

My summer research project (“Data Enabled Building Energy Savings- D-E-BES”), funded by Microsoft Azure for Research Program, was very successful. In fact, the results will be published in the upcoming edition of the Proceedings of IEEE, Special Issue on Smart Cities. I’m the lead author on the paper, which is a great honor for me.

In this research, we wanted to take advantage of two key factors; the difference in the internal temperature of apartments that are located on different floors in a residential building, and the difference in the temperatures that make apartment residents feel comfortable. Our team used two years of data from smart sensors that were installed in 310 apartments in New York City.  We were able to demonstrate that average energy savings of as much as 28 percent are possible by assigning residents to apartments whose free-running temperatures (i.e. the temperature of an apartment when no heating or cooling is on) are closer to the residents’ thermal-preference temperatures (i.e. the apartment’s temperature after a resident has turned on heating or cooling using the thermostat control).

Can you give an example of how you assigned residents to different apartments?

So this is where the data science tool kit comes in handy. We first tried to understand what factors affect the internal temperature of an apartment and the factors that affect a resident’s thermal-preference temperature. Previous research in this field had highlighted the importance of outside temperature, relative humidity and time of day. We demonstrated that the orientation (eg. north, south, east or west) and floor-level of the apartment also significantly affect both an apartment’s internal temperature and a resident’s preferred temperature. Then, we created two machine-learning models. The first predicted the internal temperature of an apartment at a given time, outside of temperature and relative humidity. The second model predicted what temperature a specific resident would feel comfortable at in a particular apartment. Finally, an algorithm takes in these predictions for the summer months and provides us with favorable apartment-resident pairings that lead to maximum energy savings in the residential building.  

So what are the practical applications of this approach?

The practical application of this idea is in community housing and dormitory-style housing. Our framework is completely non-obtrusive, meaning that it provides data-driven feedback to building managers. The building manager can then use this feedback for optimal assignments. For organizations such as Columbia Faculty Housing, which provides housing for more than 40 percent of Columbia University members, our energy-saving framework (D-E-BES) is an opportunity to conserve energy in the university dorms, where residents are seldom responsible for their utility bills and therefore have little incentive to conserve energy. Since residence housing is provided on a semester-long basis, the Faculty Housing Division does plan to use our D-E-BES framework for assigning apartments to residents based on a student or faculty member’s previous thermal history.

You also had a successful Capstone project, which all DSI students must complete in their final semester. Can you discuss that project?

After my research, it was the second best experience for me at DSI. Our project focussed on the retail industry. We worked with Synergic Partners and used purchase data of Instacart customers to create an online web application for marketers. The web application provides predictions on when a customer will make the next order. It can also predict the contents of the order, personalized recommendations for the customer as well as customer segmentations that could be used in marketing campaigns. Working on this project gave us a complete understanding of the responsibilities of a data scientist. With tight deadlines and delivery expectations, we learned the importance of teamwork, work delegation and how to communicate our results to audiences from different backgrounds. It also gave us a chance to implement everything we had learnt in our classes and made us confident that we were ready to enter the workforce as data scientists.  

What were your favorite classes?  

Applied Machine Learning with Professor Andreas Mueller and Bayesian Models for Machine learning with Professor John Paisley.  I listed those classes on my resume and recruiters often ask me about them. In fact, I remember having lengthy discussions on the applications of Bayesian modeling in all of my interviews. Recruiters really appreciate it when you can comfortably explain the math behind machine learning.

Did you learn a lot from your fellow students, who come from all over the world to study at DSI?

In today’s world, where opinions about peoples and cultures are often formed only from media reports, being in a place like DSI/Columbia was a blessing. My friend circle included students from Iran, Turkey, Cameroon, Ghana, Egypt, Canada, China, Indonesia, France, as well as from all over the United States and India. I learned that there is little that separates us but much that unites us – our understanding of the world, its many problems, and our dreams to use data science to forge a better future.

Are you glad you took a chance and left a good job to study at DSI? In the end was it worth it?

At DSI, I had the opportunity to work on research projects with prominent professors in an area I love: smart cities. I had some great classes and with terrific professors, many of whom I worked with very closely. I made good friends from all over the world and had the chance to experience living in Manhattan. I now possess the most coveted skill set (data science) on the market and have a master’s degree from the world’s leading data science institute. And in a week I’ll begin working for IBM in a job I’m certain to enjoy. So yes – it was totally worth it.