At Auric AI Labs, I detect and track military objects in high-resolution satellite imagery. I built synthetic satellite data in Blender and ensembled detections with formation and size priors, improving precision by 10.67% and recall by 12.40%.
At Samsung Research Institute, Bangalore (SRIB), I designed a deep learning model for joint video deblurring and stabilization on edge devices. I integrated IMU-guided fusion and temporal modeling, reaching a PSNR of 39 and stabilization score of 0.893 while outperforming SOTA benchmarks with reduced computation.
My research includes FlowIID, which reduced intrinsic image decomposition error by 25% over previous SOTA, and work on low-light visual SLAM that increased feature retention by 80% at 100 lux. I’m pursuing an M.Tech. in Artificial Intelligence at IIT Gandhinagar, where I’ve also been a Teaching Assistant.

