About Me

I am a Ph.D. student in the department of Electrical and Computer Engineering at the University of California San Diego, where I am working under the supervision of Prof. Henrik I. Christensen in Autonomous Vehicle Lab (AVL). Before starting my Ph.D. program at UCSD, I was an undergraduate in the department of Electrical and Computer Engineering at the University of Tehran. I received my bachelor’s degree in Electrical Engineering with a minor degree in Computer Engineering. I was fortunate to be advised by Prof. Hamed Kebraei during my undergrad studies.

Research Interests

I am passionate about exploring cutting-edge solutions at the intersection of machine learning, computer vision, and robotics. My research primarily focuses on how we can model and represent complex urban environments more efficiently and accurately, with applications ranging from autonomous driving to intelligent systems.

Currently, I am working on 3D Gaussian Splatting for modeling urban environments. This technique involves representing large-scale, dynamic scenes using 3D Gaussian functions to render high-fidelity representations. The work has significant applications in autonomous driving and robotic perception, enabling more efficient scene understanding for tasks like lane detection, topological reasoning, and trajectory prediction. By advancing how spatial information is encoded and processed, my goal is to improve the accuracy and efficiency of perception systems in real-world, urban environments.

In addition, I am deeply invested in solving the challenges of scalable map perception, especially in the context of driving environments. One of the major objectives of my work is to leverage accessible data sources like standard-definition (SD) maps and satellite imagery to create robust representations of road networks and lane topology. This includes developing models that generalize well across diverse sensor configurations and geographic regions, enabling improved perception and decision-making in autonomous systems. I also focus on using large-scale, geo-referenced high-definition (HD) maps to train scalable models that can enhance scene understanding without the heavy reliance on expensive, sensor-specific data.

I am also deeply engaged in uncertainty estimation in machine learning models, aiming to develop probabilistic methods that make these models not only more accurate but also more reliable in real-world environments, where dynamic and unpredictable conditions are prevalent.

Previously, I worked on developing scalable computational algorithms for semidefinite programs (SDPs), focusing on optimization techniques such as spectral bundle methods and chordal decomposition. These methods addressed the inherent sparsity in SDPs and were crucial for improving scalability without adding extra variables. My work in this area contributed to advancements in control theory and optimization, pushing the boundaries of solving large-scale SDPs efficiently. This foundational research continues to inform my approach to scalable algorithms and robust systems.