Research scientist working on 3D computer vision, geometry, reconstruction, and autonomous systems.
My work explores how machines can recover structure, pose, and geometry from visual observations—especially when the available views are sparse, noisy, or far apart.
- MVS2D — efficient multi-view stereo with attention-driven 2D convolutions. Paper
- RelativePose — extreme relative-pose estimation for RGB-D scans through scene completion. Paper
- FvOR — robust joint shape and pose optimization for few-view object reconstruction. Paper
- HM3D-ABO — a photorealistic, object-centric multi-view dataset for 3D reconstruction. Paper
Starhaven is my evolving personal computing environment: dotfiles, macOS services, devices, sounds, and automations designed to make a workspace feel thoughtful and alive.

