About
I'm a second year Ph.D. student in Computer Engineering at the University of Maryland, College Park, working under the supervision of Professor Ming Lin. Previously, I got my B.S. in Computer Engineering from Sharif University of Technology, where I worked with Professor Mahdieh Soleymani, and Professor Mohammad Hossein Rohban.
My research focuses on making AI systems more reliable and safe, especially when they are deployed in real-world, resource-constrained settings. A recurring theme in my work is understanding when models appear accurate on average but behave unexpectedly in important or edge cases. My current focus is understanding how quantization can change model decisions, not just overall accuracy, particularly under distribution shift and rare cases. More broadly, I investigate efficiency and safety trade-offs in learning-based systems.
Publications & Preprints
- Quantifying and Modeling Driving Styles in Trajectory Forecasting IROS 2025
- Decompose-and-Compose: A Compositional Approach to Mitigating Spurious Correlation CVPR 2024
- Annotation-Free Group Robustness via Loss-Based Resampling OOD-CV at ICCV 2023
* denotes equal contribution.
Work Experiences
TAPSI is an Iranian ride‑hailing company similar to Uber, and stands as one of Iran’s largest and most technologically advanced companies. Notable projects I contributed to include:
- Estimated Time of Arrival (ETA): Investigated and developed traffic prediction models, improving ETA accuracy using techniques inspired by industry best practices.
- Location Search Engine: Processed raw data and built an in-house search engine with optimized offline and online metrics.
- GPS Denoising: Applied noise reduction techniques to enhance GPS data quality and reliability for location-based services.
- Destination Suggestion: Built ML models to recommend destinations based on user ride data.