PhD in ECE at UCSD
I am a second-year PhD student at UC San Diego advised by Prof Dinesh Bharadia. I work at the WCSNG Lab, where my research focuses on developing scalable deep learning systems that perceive, reason, and interact with the physical world, with applications in autonomous driving and embodied intelligence. My interests lie at the intersection of multi-modal perception, sensor fusion, and large language models (LLMs).
I completed my undergraduate at IIT Hyderabad, in the department of Electrical Engineering with a minor in Computer Science. I was fortunate to be advised by Prof. Aditya Siripuram and Prof. Ayon Borthakur. My projects spanned data inpainting, graph learning, efficient algorithms for neural network training, and image processing.
Besides work, I enjoy swimming, squash, badminton and ultimate frisbee. I love listening to music and my favourite activity to unwind is to go on long drives. I also represented my squash team at multiple competitions in IIT Hyderabad.
I am broadly interested in the intersection of sensing and machine learning. I am very passionate about developing deep learning systems that have real-world impact and drive innovation.
We present C-Shenron, a radar simulation framework integrated into CARLA that generates realistic radar measurements by fusing LiDAR and camera data. Our framework supports configurable radar parameters and demonstrates that radar-camera fusion models achieve performance equivalent to LiDAR-camera baselines on CARLA leaderboard metrics.
C-Shenron is the first realistic radar simulator that utilizes LiDAR and camera sensors to generate high-fidelity radar ADC measurements from physics-based modeling. We demonstrate improved performance in end-to-end driving scenarios and aim to rekindle interest in radar-based self-driving research.
We propose SPELA, a learning algorithm that operates with local loss functions using a single forward pass, significantly reducing memory and computational requirements. SPELA eliminates gradient propagation while demonstrating competitive performance with backpropagation on standard benchmarks, making it ideal for resource-constrained edge AI applications.
We propose an algorithm for simultaneous data restoration and graph structure learning using an interpretable neural network designed from unrolling framework. Our closed-loop feedback mechanism guides graph learning based on inpainting performance, achieving robust results under noisy conditions on sensor network datasets.
Throughout my academic journey, I have gained valuable industry and research experience through internships and leadership roles.
Designed and implemented a cross-talk cancellation algorithm for spatial audio playback. Developed an efficient low-order filter approximation using bi-quad and all-pass filters.
With Prof. Aditya Siripuram
Assistant for Linear Systems & Signal Processing and Probability for AI. Conducted tutorial sessions and formulated questions for exams.