I am a deep learning researcher, pursuing my PhD in CS@Stanford. I am currently exploring ideas on designing computer vision systems motivated by human behavior and thought.
Previously, I worked as a Research Engineer at Avataar, where I built solutions for image-to-3D reconstruction problem. Before that, I was a Pre-Doctoral Researcher at Google Research India. I worked towards solving the problem of predicting super-resolved segmentation maps from satellite imagery in the Earth Observation Science team under the guidance of Dr. Varun Gulshan and Alex Wilson.
Before that, I graduated from the Indian Institute of Science Bengaluru (IISc) with an M.Tech. (Research) in Computational and Data Sciences. I was part of Video Analytics Lab (VAL), where I worked towards my thesis titled Landmark Estimation and Image Synthesis Guidance using Self-Supervised Networks, advised by Prof. R. Venkatesh Babu and Dr. Varun Jampani.
I have been fortunate to have beeen part of Google Summer of Code twice to contribute to the JuliaLang under the mentorship of Mike Innes. In the summer of 2019, I built Duckietown.jl, a differentiable self-driving car simulation environment that enables the training of agents using differentiable programming. While in the summer of 2018, I contributed to Flux’s model zoo with Reinforcement Learning algorithms and built AlphaGo.jl to train and play zero-sum board games using Alpha(Go)Zero algorithm in JuliaLang.
Email / CV / Google Scholar / GitHub / LinkedIn
September, 2024 : Started PhD in CS at Stanford University.
August 2024 : Our work AttributeDiffusion, which generates diverse-attributed image, is accepted at WACV 2025!
July 2024 : Our work PreciseControl, which combines best of StyleGAN and Text-to=Image Diffusion Models to generate diverse images with precise attributes is now accepted at ECCV 2024!
November, 2023 : Serving as a reviewer in CVPR 2024.
October, 2023 : Started as a Research Engineer at Avataar!
September, 2023 : Serving as a reviewer in ICLR 2024.
May, 2023 : Serving as a reviewer in NeurIPS 2023.
February, 2023 : Our work NoisyTwins, which aims at long-tail and few-shot image generation for StyleGANs, got accepted in CVPR 2023!.
November, 2022 : Serving as a reviewer in CVPR 2023.