AR Data Sciences
Guest Column: Debanjan Mahata, Ph.D.
Updated: Jul 24, 2019
ARKANSAS GRAD MAKES GOOD
Unfortunately, he had to leave the state to do it
Debanjan Mahata, Ph.D,
Research Scientist, Bloomberg
FOUR YEARS AGO, I was just receiving my doctorate from UA Little Rock’s Donaghey College of Engineering and Information Technology (EIT), and like most young tech grads, I had plans. This is the story of how those plans did, and didn’t, work out.
I was born in Calcutta (now Kolkata), in a well-to-do family. My father was a police officer, an athlete, a star rugby player, a polo player, a flower enthusiast, and a history buff. My mother was a schoolteacher who was the main pillar of the family; she took care of all the nitty-gritty family details. I consider myself extremely lucky to be raised by such parents. I take them both as my inspiration, and I continuously strive to be disciplined and hardworking like them.
My elder brother and I grew up with all the Indian family values. I went to a very good English medium school, which was a privilege. After my schooling I went to Banaras Hindu University, which always ranks among the top universities in India, for my undergraduate and graduate degree. I didn’t have firm goals, except to become a good human being and embrace simplicity. My career goals evolved with time and circumstances.
Initially, I was interested in pursuing an undergraduate degree in chemistry and had started to focus on that track. But my parents foresaw the future and guided me toward computer science, as that was also an option. I think my love for chemistry got manifested through the love of my life—yes, I am married to a chemist!
As I learned computer science, my initial interest was in the fundamental working of the Operating Systems and Computer Networks. That gradually changed as I took more courses. Finally, when I took courses related to Artificial Intelligence and Soft-computing, which covered some basic topics of machine learning, I became attracted to that. Since then I have been a student of machine learning and still continue to learn its various aspects, and to apply them in my work.
I arrived in Arkansas on July 30, 2011, and started my Ph.D studies in mid-August of that year. But my coming to Arkansas was a happenstance; when I was finishing my higher education in India and starting to plan toward a Ph.D, I had never even heard of Arkansas. I was closely following the research done by Dr. Huan Liu at Arizona State University, and Nitin Agarwal, who was then a Ph.D student supervised by Dr. Liu. They were working on topics like mining data from the blogosphere, and had published an algorithm for identifying influential bloggers. I was very influenced by their research, as I was working on a very similar topic for my final semester project in India. Therefore, I wrote an email to Dr. Liu, expressing my interest and asked if I could join his lab.
Unfortunately, at that very moment Dr. Liu didn’t have funding and he recommended that I join the lab of Dr. Agarwal, who by then had received his degree and taken an Assistant Professor position at University of Arkansas at Little Rock. I immediately contacted Dr. Agarwal. Interestingly, I did eventually get an offer from ASU but chose UALR instead, as I now had the exact topic that I wanted to work on. My main goal was to work with someone whose research interests were well aligned with my desire to study the content produced on real-life events in the blogosphere, particularly the citizen-journalistic sources of information.
I liked UA Little Rock. It’s a smaller university in comparison to many other universities in the United States, but that worked in my favor. I felt that I got more attention from the professors, and it was very easy to collaborate with them and learn about all sorts of interesting topics. All the faculty members worked hard to help the students be successful.
As a graduate student, I had a very good time at UA Little Rock. However, my main focus was on my research instead of the courses that I took. The emphasis of the courses leaned toward the applied aspects of the subjects. Some of the courses took real-life use cases and showed how to apply the techniques taught in the class. I really enjoyed all of them, especially Dr. Nitin Agarwal’s lectures on Social Computing, Dr. Ningning Wu’s course on Data Mining, Dr. Daniel Berleant’s course on Information Theory: Principles and Theory, and Dr. John Talburt’s course on Entity Resolution and Information Quality. In my career path so far, the techniques learned from these courses have been most useful.
I would say, however, that UALR needs to recruit faculty members who can teach core machine learning theory, statistical analysis, calculus, text mining, and natural language processing. There should be courses at the undergraduate level dedicated to teaching big data tools. During my tenure at UA Little Rock, there was no emphasis on deep learning at all.
As I progressed in my studies, my goals began to change. Initially my aim was to join academia. But with the sudden boom in the area of machine learning and the data scientist being the sexiest job in the 21st century, I couldn’t help being attracted to industry, where there were genuine machine learning problems to be solved. For the first time, researchers in computer science had access to humungous volumes of data and the necessary hardware and software to process them. This was certainly a momentous time for mankind, and I wanted to be part of this revolution in which all kinds of industry was going through a paradigm shift.
Not only that, but there was great demand for my skills in industry, from which I received much more attractive offers than what I could have got from academia. The positions available in industry also required candidates to do research, while providing the resources to do so. Joining industry was certainly looking like a win-win situation.
By this time, I was married. My wife had already received an admission offer from UA Little Rock, and she was still in India waiting for her visa to be processed. The initial plan was that I would finish my Ph.D and get a job in Little Rock, and my wife would pursue her Ph.D studies in chemistry. My doctorate was under UALR’S Integrated Computing program, and my major area of focus was Information Quality. My job interests were in the area of applied machine learning, particularly in the area of natural language processing or computational linguistics. I hadn’t taken any courses on this topic, as none was available at UALR. But while working toward my degree the interest had grown, as I handled more and more social media data curated both by humans and bots. The computational approaches to parse and understand colloquial content written by people in different languages really interested me, and I planned to work in the area of natural language processing. I confess to being very picky and interested in only those jobs, which had an emphasis on research.
I was actually planning to stay in Arkansas if I received a good offer or if any major company approached me. I love nature and feel better and more energized when I am surrounded by it, so Arkansas, the “natural state,” was a natural choice for me. We planned to take an apartment by the river so I could walk to the Big Dam Bridge. That never happened, however. I never was asked to interview with any company in Arkansas. I did apply to Acxiom for a Data Scientist position, but they didn’t respond. I had also submitted my resume to the campus recruiter in UA Little Rock, for positions available at Acxiom, and the results were the same. I did consider Black Oak Analytics (which no longer exists), as my Ph.D advisor was the Chief Scientist of the company. Meanwhile, I was getting calls from companies in Silicon Valley. Unable to stay in Arkansas to do my work, I accepted an offer from Infosys Ltd, in Palo Alto. This meant that my wife, who had moved to Arkansas and worked one year on her Ph.D, had to interrupt her studies so we could move to California.
We were there two years, and then I received a very nice offer from Bloomberg in New York, where I am now a Research Scientist. My role at Bloomberg allows me to work at the intersection of natural language processing (NLP), information retrieval, machine learning, and software engineering. I am not only responsible for researching some of the challenging problems related to these areas, but also to build real-world solutions around them. The resulting applications ship as products in the Bloomberg Terminal, enabling our clients around the globe to make smarter, more informed decisions about their business and financial strategies. It’s the perfect job for me, and my wife has renewed her Ph.D studies at the New Jersey Institute of Technology. Also, we’ve been blessed with twin babies, a son and a daughter. Life is going well for us!
I don’t want to make too much of my not receiving interview requests in Arkansas. That was four years ago, an eternity in technology terms, and I was interested in very specific kinds of work. As far as I remember, in those days technocrats did fully understand the value of data and data-driven products, but there was no such platform in Arkansas—except Acxiom—that might have attracted me to work on interesting problems.
I hope, for the sake of Arkansas tech graduates to come, that that situation is changing. Today I’m happy to be receiving the ACDS newsletters and to read about the exciting new developments in my old “natural state.” I recently read the Scott Spradley interview and really loved it—I mostly agree with what Scott had to say and his vision. I also read Mike Preston’s guest column, about all the new companies and tech jobs coming to Arkansas. Given that these companies are coming, there should be a great push in the University programs to train the necessary workforce in order to retain these companies and to create centers of excellence in research and development in Arkansas.
It is true that today most companies are tech companies, or else they’ll be left behind. But even more important is: Are they data driven? Simply changing the operation from a brick-and-mortar structure to a technological stack isn’t good enough. Today’s big tech companies are actually fueled by data and driven by the decisions that they make on top of it. Most of their efforts are toward organizing and managing high-quality data and building the infrastructure to gain insights from that data; for some, their effort is to build full-fledged revenue-generating applications using this data.
Apart from the sectors mentioned in the articles of the ACDS newsletter, I think there is a huge chance to develop indigenous companies focused toward improving the agricultural yield of the state. Agriculture is Arkansas' largest industry, adding around $16 billion to the state's economy annually. Just as Scott Spradley is taking Tyson to the Cloud, I think Arkansas’ agriculture can also be taken to the Cloud, with digital farming being an emerging technology. In fact, the entire sector can benefit from machine-learning-based automated systems. A good example of a company working on similar things would be Landing AI. I would like to see ACDS work with such a company to promote AI and to transform the workforce.
Another area ripe for AI transformation is medical sciences. I believe there is a huge opportunity at places like UAMS to develop AI-based bio-medical systems, both for the patients as well as the clinicians. AI in health care is going to be a huge industry, one crucial to providing better service to the patients, doctors, and clinicians. Artificial intelligence is already disrupting the healthcare industry, and UAMS has the necessary platform to develop AI-based systems, as well as the capacity to modernize their curriculums and introduce new courses. They just need to get going on embracing AI, and also to find the will to recruit faculty members that can drive the organization toward such efforts.
I have good memories of Arkansas and would be very happy to play a part in the state’s Data Science initiatives, and to help in any way that I can. For example, in addition to my work at Bloomberg, I’m also a member of the adjunct faculty at Indraprastha Institute of Information in Delhi, India, where I advise and collaborate with both undergraduate and graduate students working at MIDAS lab on topics related to natural language processing. Some of the broad areas in which I work are summarization, keyword extraction, social media mining, and sentiment analysis. This is a voluntary position that allows me to be in touch with academia, which I greatly enjoy.
With my current set-up and family, I’m not sure I would think about actually returning to Arkansas. But who knows about the future? Say there is a time when I have an option to return, and my family supports me. In that case, I would like to see myself in a leadership role in the area of education, especially in the area of Data Science focusing on fundamental and applied research in the area of natural language processing and machine learning. In the meantime, I look forward to continuing this dialogue!