I am currently working as a Research Associate at Adobe. I completed my Integrated M.Tech in Mathematics and Computing at the Indian Institute of Technology (IIT) Dhanbad. Since my sophomore year, I have been actively involved in machine learning research, focusing on efficient machine learning, multimodality, and reinforcement learning, and have been fortunate to publish my work in top machine learning venues.
In addition to research, I enjoy participating in Kaggle competitions and in my free time, I like to paint and play badminton.
Experiences
Worked on improvement of reasoning process of smaller language models on mathematical tasks.
Worked on alignment of large language models with user preferences by using reinforcement learning for synthetic query generation.
Developed a novel, efficient multimodal model for user interface navigation, achieving state-of-the-art performance on the largest UI benchmark, Android In The Wild (AITW).
Developed MixBin, a binarization-based network compression method combining binary and full-precision components in CNNs, outperforming previous methods on ImageNet.