Soumyo Chakraborty, WG’01, has class. When I remind him that he was the top-ranked student in our class for the two years we shared at Wharton, he protests, “No, no, there were five of us. I think they called them Palmer Scholars — I was just one of them.” Despite his wanting to be considered an ordinary man, as WCNY’s new Chief Technology Officer, Soumyo is exceptional.
Soumyo came to Wharton, having worked as a software engineer at a tech startup and going deep into computer science at ATT Bell Labs. His goal in attending Wharton was to broaden his perspective by learning about finance, marketing and management. Wharton did open his eyes to see things from various angles. And he transitioned into roles, including engineering management, project and product management, and founder and CEO of startups, that required him to incorporate those skills.
What did you apply from Wharton to your work?
I had a class with Professor Jitendra Singh, in which I wrote my first business plan. It was for a bioinformatics business, applying IT to biology to make drug discovery more efficient. I learned to appreciate the business aspects of a competitive landscape. After that, I wrote numerous business plans, outside of class, for my company, which we used to incubate new business ideas and help them raise capital.
Within a few years, I decided to try my hand at running my own firm.
None of my ventures became household names. Some went, at best, sideways. One area that fascinated me, and still intrigues me, was integrating renewable resources into the electric grid. Along with my background in computer science, I applied what I learned in finance and operations management to this issue. One question was, “As battery-dependent vehicles increase their demand on the grid, along with the somewhat unpredictable timing of when they get plugged in, what would be the overall effect? Will it destabilize the grid?” I tried to use computer models and simulations to understand and analyze the impact, but I could not make money on it.
Since then, I have made it a personal research interest and have been periodically publishing papers at academic conferences on the topic. Currently, I’m working on a publication in which I analyze electric power hubs, and present a way to model the behavior of bulk power prices. As we obtain more energy from solar, does pricing become more volatile based on weather variables like solar irradiance (the power per unit area received from the sun in the form of electromagnetic radiation)? There are 20 different factors I’m analyzing that can affect electricity pricing. You can explore these types of questions with the use of machine learning models.
And then you came to Two Sigma, one of the world’s top performing hedge funds.
Eight years ago, Two Sigma was a much smaller firm. They liked that I had such a variety of interests, and a hybrid background combining technology and finance. Due to Two Sigma’s culture of learning, they also were supportive of my research interest in the science of the solar industry and publishing of papers. I had two children growing and thought that I better get back into the job market, and I joined Two Sigma. Today, we have several different investment business lines, all with a heavy focus on engineering and technological innovations, and with the primary external products being funds. I work in the engineering division, focusing on the management of technical projects and products.
What can you say about the hedge fund industry in general?
In the past five to 10 years, there has been an explosion of computer power and an explosion of the availability of data that provides the essential fuel to drive algorithms spanning multiple industries. We see this rapid growth in the power of computer algorithms in a wide variety of areas like speech recognition, language translation, financial modeling, self-driving cars or recommendation systems such as Pandora. Most of our problems in corporate and daily life can be structured as pattern-matching problems, which can be solved by appropriate datasets and computer algorithms. That makes machine learning and data science an exciting area.
Can artificial intelligence learn how to build models through machine learning?
Let’s say you’re analyzing handwriting. There are training sets — say, sets of A’s handwritten by different people. The computer tries to learn the patterns that represent A. That learning is embedded into models.
Then, take the model and give it something it has not seen before. The model tries to guess what it is. The more training data you can supply to the model, the more it can improve its ability to predict. In the old days, we lacked data. But today, we have so many sensors — providing so much data to train our models. There are also vast amounts of human-generated data easily available today. For example, how do you make a translation machine? You need a huge amount of translation data to train the machine. One source of translation data is the transcripts from the United Nations and EU proceedings and speeches that are translated by professional human translators into many different languages. Those are now available to train the computer models.
This raises the need to store the data and to run computations on the data. So, the growth of data, cheap data storage and compute power has driven the progress in machine learning.
What are your plans as the Chief Technology Officer of WCNY?
I’d like to leverage technology to improve member engagement, which is also a key area of focus of WCNY’s President and the leadership team. One way is to create cloud-based webinars, allowing members to join our events remotely or view events afterward. We will have a camera at the event (a laptop camera is good enough), and will make the event interactive. A remote participant can ask questions and listen to answers. We evaluated a few webinar vendors and made a selection.
A second initiative has been to improve collaboration among volunteers using the cloud-based Google G Suite platform. This will provide shared spaces, where Club members can co-create and share documents in a secure way. It will also enable us to build automated workflows. One example is a G Suite app that we implemented for our finance team. Team members can use it to set up rule-based approval workflows to reimburse people more efficiently.
My third effort is to improve the integrity of our data. As you know, this initiative will affect the magazine, and potentially help address issues we faced in the past where some members were accidentally removed from the mailing list. I received a data dump from the school, and ran initial analytics involving data from both the school and Club databases. I am still studying this and working with our communications team to come up with possible solutions.