About

I’m a second year Ph.D. student in Electrical and Computer Engineering at the University of Maryland, College Park, working under the supervision of Professor Ang Li in the CASE Lab. Previously, I got my B.S. in Electrical Engineering from Sharif University of Technology.

My research focuses on efficient and robust foundation models. I study how architecture, data, modalities, and computational constraints shape the behavior of large models, with the goal of understanding how they can be compressed, adapted, and deployed without sacrificing their core capabilities.

Building on these insights, I develop methods for model compression, quantization, robustness, and parameter-efficient adaptation, especially for multimodal systems. I am interested in making large models practical under real-world constraints such as limited memory, limited computation, distribution shifts, and deployment on edge or robotic platforms.

Broadly, my goal is to bridge model understanding, algorithm design, and efficient deployment, enabling AI systems that are not only powerful, but also reliable, economical, and usable in real-world scientific and robotic applications.

Publications

Recti-Q: Feature-Space Rectification for OOD-Robust Quantized Perception in Edge Robotics
H. Yaghoubi*, P. Pilevar*, M. Lin (IROS 2026).

Identifies and quantifies the OOD robustness gap introduced by 4-bit PTQ in large vision backbones on resource-constrained robotic platforms, and proposes a frozen-backbone LoRA adapter trained on source data only that preserves PTQ memory savings while closing most of the robustness gap.