I am a Senior Engineer/Researcher at Qualcomm developing low-power and compute-efficient computer vision/machine learning algorithms.
Before joining Qualcomm, I was a postdoctoral researcher at Northeastern University working with Prof. Jennifer Dy, Prof. Octavia Camps and Prof. Dana H Brooks. I completed my PhD in Computing and Information Sciences from Rochester Institute of Technology under advisement of Prof. Linwei Wang.
Research
I feel glad to have worked in diverse areas like signal processing, medical imaging, deep learning, learning theory, kernel methods, generative modeling, reliable estimation, and score-based generative models during my undergrad, PhD and Postdoc. The richness, diversity and cross-fertilization of ideas from different fields like Physics, economics, mathematics, computer science, statistics, electrical engineering, control and neural network architecture engineering is exciting and truly satisfying in modern AI and machine learning. Ideas have cross fields and found interesting adaptations and interpretations resulting into ingenious solutions of problems both in machine learning and outside. It is beautiful. I feel happy to have contributed in this process. During PhD, I applied VAE as inductive bias in solving inverse problems (see here), and connected generalization with smoothness and invariance (see here). Later, I applied idea from learning theory to KL divergence estimation (see here), and introduced the geometric interpretation of score-based generative models.
At Qualcomm, I am excited to work on machine learning, computer vision and AI research works that will be realized as product, and will have real-world impact on everyday lives of numerous people.
Resume PhD ThesisPhD in Computing and Information Science, 2020
Rochester Institute of Technology
BE in Electronics and Communication Engineering, 2012
Institute of Engineering
Can we learn meaningful representation from real world videos or biomedical signals ?
How to understand and improve generalization and robustness in deep networks?
Apply geometric information while solving inverse problem.
Apply probabilistic modeling, deep generative modeling and inference to solve inverse problem.
Apply smoothness based regularization to help semi supervised learning.
SOME THOUGHTS