United States
I’m a fourth-year Ph.D. student in Computer Science at UC San Diego, advised by Prof. Tzu-Mao Li, working on 3D reconstruction, neural fields, inverse rendering, and computational imaging. Actively looking for summer internship positions! My work spans algorithms and hardware: during my B.S. and M.S. at KAIST I led the design and construction of two generations of large-scale light stages with over a hundred cameras and hundreds of polarized light sources, including custom optics, electronics, and distributed capture software for high-fidelity multiview 3D scanning. On the computational side, my works include Spectral Prefiltering of Neural Fields (SIGGRAPH Asia 2025), which introduces analytically prefiltered neural fields for single-pass filtering of signals under anisotropic filters such as Gaussian, Box, and Lanczos; FIPT: Factorized Inverse Path Tracing (ICCV 2023), an indoor inverse rendering algorithm; and DeepFormableTag (SIGGRAPH 2021), a learned deformable fiducial marker system trained end-to-end with photorealistic differentiable rendering. Building on this background, my recent projects explore generative models and vision-language models for part-aware reconstruction. I use VLMs to reason about segmentation, materials, and geometry, and to connect image-level cues to 3D assets. My current goal is to develop practical pipelines that combine this semantic guidance with physically grounded reconstruction and rendering.
Worked on a novel immersive application on Vision Pro. My work aims to extend the Spatial Photos feature.
I built and extended the language processing packages of the firm through modifying and contributing open-source project flair • I preprocessed finance related news dataset for the Named Entity Recognition(NER) and Entity Relation Extraction(RE) tasks • I trained a NER model that detects company entities, and improved accuracy significantly • I developed a brand new RE model that returns entity relation graph for given sentence • I developed an alpha for stock investment using textual finance data and stock market information