Undergraduate major: BE Computer Science
Undergraduate Institution: BITS-Pilani
Hometown: Noeda, Inda
After 4 years of working in industry, Sugam comes to the NLP MS program with a drive to learn about NLP theory and how to build NLP models. As a Student Ambassador, he brings people together, socially and academically, and hopes to eventually apply his NLP skills to building user-centered applications in languages other than English.
What led you to your interest in NLP?
I did my internship in my last year of undergrad at Samsung R&D, Bangalore. My intern project was working on a machine learning-based feature for mobile phones. After that internship, I really understood that the industry is slowly moving towards machine learning, and that I needed a theoretical background as well.
After I joined my full time role at Samsung R&D, my primary job was to put a language model on a mobile phone. Then I got introduced to different text-classification paradigms, and then transformers. I eventually moved onto another project on semantic search at a different company. That was 2020. And then NLP was taking off: GPT-3 was released and the buzz was high. But these large NLP models were still not in the hands of actual users.
I kept thinking about one thing: I know how to use a lot of models, I know how to read papers, but I don’t know what it takes to build them. I didn’t really understand why things are the way they are, and I realized that I needed to go back to academia to study from experts to understand the natural progression of NLP. The NLP MS program structures it for you, provides you with guidance, and answers so many questions for you which will otherwise take days to solve.
What do you think about the program so far?
Even though I had four years experience using, developing, and studying NLP applications, the NLP 201 series kind of took me by surprise by how little I knew about NLP prior to the deep learning era. There were a lot of things which Professor Jeff Flanigan explained that I still understand because of the manner in which he explains things. I don’t think I will forget the basic concepts that he taught. I really appreciate the way he explains nuances in NLP and how they fit into today’s GPT-era models.
A second experience is from NLP 244. I follow NLP stuff on Twitter, and a paper came out in January. During our next class, Professor Delip Rao was talking about the paper that was released just two weeks ago. He explained the paper and put it into a framework and the entire course itself. I don’t know if there’s any program in the world that has a two-week turnaround time.
Can you tell me about your Student Ambassador experience?
I wanted not just academic excellence, but I also wanted to contribute and find a sense of belonging and create a community. I joined the Student Ambassador Program, and I joined the Community Organizer team. We organized hang-outs like movie nights and dinners, and I made friends which I hope will stay beyond the program. I got to know people outside of academic interactions. It is really fun, and I hope to continue doing that.
What about the NLP paper reading group?
This is pretty close to my heart. Back when I was working at Samsung, there was probably not a single university that was offering a dedicated single course for NLP let alone an entire program. It was difficult to understand the developments in NLP. I decided to host a monthly paper session which became a community-oriented thing. We had volunteers every month speak about the things they were working on. This was another way of team bonding. I started a knowledge-sharing group immediately after joining my next company as well.
When I came to the NLP program, I noticed that there’s a new paper every two weeks. The NLP MS curriculum teaches the conceptual arguments made in papers, but in order to learn about the nuances in execution, one has to read the paper itself. That is where the success lies. I approached the Ambassador team with the idea of a paper reading group, and everyone was interested. It has been great, and I hope the program continues to do it.
What do you plan to do when you graduate?
NLP technology is not in the hands of users. It’s benefiting people who were already part of the AI world or the computer world, but it doesn’t reach people beyond the existing digital population. I want to work with companies that are trying to get more people using AI products, so that actual people can benefit.
Are there technical or societal problems that you hope to solve in the future?
The first problem I want to solve is the language problem. I come from a land of 100+ languages. I am not familiar with any AI model that is working on a large scale with regional languages in India. My mother uses her phone, but she doesn’t use any utility or productivity apps because she is not really familiar with English. My goal is to help build apps which can bring people like my mother to use AI products because they offer tremendous value. It’s my dream to move towards an inclusive language ecosystem.
The second thing that is close to my heart is educating more and more people. I love teaching. Even here at UCSC, I’m a tutor for a web application course, and yesterday I received an email from a student saying that the suggestions I gave them helped them finish their assignment on time. It’s the most wonderful feeling. I could honestly say that it’s more important than anything else to me, to see people achieve the best they can.
In India, I see a gap in the quality of education that I received versus the quality of education that most people are going to get. It will be like that for at least the next 10 to 20 years, unless people change something about it. I’m seeing AI models starting to solve math and physics problems, and I just hope I can work on an idea to use those AI capabilities to scale up the quality of education.