I'm a Ph.D. student at Boston University advised by Wei-Lun (Harry) Chao.
These days, I'm excited about investigating data bottlenecks that prevent autonomous driving systems from scaling to anywhere, anytime, and anyone.
Does the way we collect, annotate, and use self-driving data really scale toward driving anywhere? For example, should we keep putting tremendous effort into annotating 3D bounding boxes for perception? My recent work explores auto-labeling from other road observers, such as vehicles already running perception (ICLR 2025) and roadside units (ECCV 2026). More broadly, I'm interested in rethinking how the autonomous driving community approaches data (NeurIPS 2026 Position Paper). I believe there should be more scalable ways to expand self-driving all around the world.
I also work on LiDAR-guided stereo depth (IROS 2026), autoregressive 3D detectors (CVPR 2026 Findings), and data generation for multi-agent collaborative perception (CVPR 2025).
I have previously interned at Waymo and LG AI Research.
I'm an (4x) outstanding reviewer for ECCV (2022, 2024, 2026) and CVPR (2026). My ECCV 2026 work also received the Best Poster Award at the CVPR 2026 DriveX Workshop.