My Story

Learning to ask questions

At school, I was once taught that “dog is the fastest animal in the world.”

That detail captures something important about where my journey began: access to good teaching and information cannot be taken for granted.

I grew up in different parts of India, including Chhattisgarh, where teaching resources and awareness of opportunities beyond school were limited. My parents placed a high value on education and gave me room to question things. Teachers and other people around me helped me find possibilities I had not known existed.

I first learned about the IIT entrance exam in tenth standard. Looking back, getting into IIT Madras was about effort, but also about that support system. I had seen capable students with far less guidance. That experience shaped my interest in education and in technology that can make knowledge more accessible.

Finding computer science

In my first semester at IIT Madras, an introductory computer science course showed me how much a person could do with a computer and an idea. I wanted to understand that more deeply.

I ranked first among the freshman cohort and changed my branch to Computer Science. Courses in machine learning, natural language processing, and deep learning gave that curiosity a direction. I became interested in systems that could learn, interpret language, and interact with people.

One early project was Algorithm Theatre, an interactive application that used animations to explain algorithms. Another was work on word embeddings: how could a model distinguish different meanings of the same word? These were ways of exploring the connection between computation and understanding.

Learning through research

Research gave me a chance to work on questions without knowing whether the proposed approach would succeed.

At IBM Research, I worked on extracting events and entities from unstructured text using shallow neural networks. In 2017, the S.N. Bose Scholars Program took me to Carnegie Mellon for a research internship. There, I investigated geometry-aware neural networks and adversarial approaches to estimating depth from images and video.

Some ideas were difficult to make work. In the CMU project, unstable GAN training limited the approach we were exploring. Understanding that limitation was part of the research.

Alongside these internships, my work at IIT Madras included image-grounded dialogue: building models that could ask questions to identify an object in an image. I also worked with a teammate on Google's YouTube-8M challenge. We experimented with hierarchical and residual LSTMs, attention, and ensembles on a large video-feature dataset. Our team finished 16th out of 655.

Those projects connected model design with the practical work of managing data, running experiments, and understanding what the results actually meant.

From research to systems people use

My industry work has taken me through enterprise search at Goldman Sachs, applied machine learning at Microsoft, quantitative research at Millennium, and research infrastructure at Bezant.

The problems have changed, but a recurring question has been how to turn an idea into something that can be tested, repeated, and used by other people.

At Microsoft, that included understanding sales workflows and building models around real business requirements. In quantitative research, it included working with limited data and combining information from different sources. At Bezant, I took a research and backtesting platform from concept to production, with reproducible workflows and structured interfaces for agent-assisted research.

I like working across that whole process: understanding the problem, building the model or system, checking the evidence, and making it usable.

Purpose and community

Education has remained part of my story beyond my own studies. I have helped students in tribal regions of Chhattisgarh prepare for college entrance exams through schools run by Ramakrishna Mission. That work connects directly to the support I received when I was younger.

I also know something about the environments in which I do my best work. When I feel a sense of purpose and a shared community, I go all in. I value a team where the work matters, questions are welcome, and people take responsibility for what they build.

That is the thread connecting my interests today: learning deeply, building useful things, and doing it with people who care about the problem.