🎓Biography
I am a researcher at the University of Maryland, College Park, focused on the intersection of computer vision, geometry processing, and geospatial data analysis. My research centers on neural representations of geospatial data, combining implicit neural modeling, topology-aware analysis, and generative frameworks to enable interpretable, continuous, and scalable representations of the physical world. I am motivated by interdisciplinary collaboration that bridges machine learning, graphics, and scientific computing.
Looking ahead, I want to carry this work toward spatial intelligence: machines that reason about the physical world through geometry rather than pixels alone. I am especially drawn to world models built on explicit 3D representations — implicit neural fields, Gaussian splats, and triangulated meshes — where the scene structure is something you can query, differentiate, and edit.
During my BEng and MPhil at Hong Kong University of Science and Technology, I was gratefully advised by Prof. Long Quan in 3D computer graphics and vision. Currently, I am gratefully advised by Prof. Leila De Floriani.
Please reach out to chat and collaborate 😆!
“Topology is precisely the mathematical discipline that allows the passage from local to global.” — René Thom
- Spatial Intelligence
- World Models with Explicit 3D Representations
- Spatial Representation Learning
- Neural Rendering
- Topological/Morphological Analysis
- Generative Model
- Vision-Language Model
- AI4Science
- Data Visualization
- Information Retrieval
PhD in Computer Science
University of Maryland, College Park
MPhil in Computer Science
Hong Kong University of Science and Technology
BEng Dual Major in Computer Science Engineering and Electronic and Computer Engineering
Hong Kong University of Science and Technology
- [Aug. 2026] Two papers accepted at ACM SIGSPATIAL 2026: ImplicitTerrainV2: Wavelet-Guided Spatially Adaptive Neural Terrain Representation as a full paper, and Rethinking Amortized Neural Representations for High-Resolution Terrain Elevation Data as a poster paper. See you @Riverside, CA in November! 🗺️
- [Aug. 2026] Our TSAS article A Parallel Scale-Space Method for Critical Features Tracking on Triangulated Irregular Networks is now published in the ACM Digital Library (ACM Transactions on Spatial Algorithms and Systems, Vol. 12, Issue 4). 📚
- [May 2026] Back at Dolby Laboratories – Advanced Technology Group (ATG) for a second summer as a PhD Research Intern – a greenfield project proposed and led by myself: authoring text and icons as native Gaussian-splatting assets inside volumetric video. 💠
- [May 2026] New preprint Rethinking Amortized Neural Representations for High-Resolution Terrain Elevation Data is out on arXiv — benchmarking amortized neural representations on terrain heightfields and proposing HUVR+SIREN for a compact terrain neural format. 🗻
- [May 2026] New preprint ImplicitTerrainV2: Wavelet-Guided Spatially Adaptive Neural Terrain Representation is out on arXiv — a follow-up to ImplicitTerrain toward a compact, efficient neural terrain data format. 🏔️
- [Mar. 2026] Our paper Geometry-Guided Camera Motion Understanding in VideoLLMs, developed during my 2025 summer internship at Dolby Laboratories – Advanced Technology Group (ATG), is now accepted @CVPR2026 Workshop PVUW. 🎬
- [Mar. 2026] Our paper SASNet: Spatially-Adaptive Sinusoidal Networks for INRs is accepted at CVPR 2026. See you @Denver! ✨
- [Aug. 2025] Completed a rewarding summer journey as a PhD Research Intern at Dolby Laboratories – Advanced Technology Group (ATG). Grateful to my mentor Harsha Musunuri and manager Guan-ming Su for their guidance and support throughout this internship. 🚀
- [Oct. 2024] Our paper scale space for TINs received the best paper runner-up at ACM SIGSPAITAL2024. 🏆
- [Aug. 2024] We have one paper accepted as Oral presentation at ACM SIGSPAITAL2024 See you @Georgia. 🎤
- [Apr. 2024] Our paper ImplicitTerrain is accepted as Oral presentation at INRV2024 (CVPR2024). See you @Seattle. 🌄
- [Aug. 2021] First day at University of Maryland, College Park, working with Prof. Leila De Floriani. 🎓
A wavelet-guided, spatially adaptive implicit neural representation for digital elevation models. A wavelet complexity field localizes high-frequency capacity to complex terrain, and post-training compression reaches 1.23 bpp — 66.25 dB PSNR on Swiss terrain tiles, +5.70 dB over prior work with 3.2x fewer parameters.
Nov 3, 2026
A controlled benchmark of amortized neural representations on high-resolution (1 m/pixel) terrain elevation data, and HUVR+SIREN — a hypernetwork with a smooth, analytically differentiable decoder that attains the best height and derivative fidelity with no extra per-tile storage.
Nov 3, 2026
A spatially-adaptive sinusoidal network for implicit neural representations. A frozen frequency embedding fixes the frequency support while jointly learned spatial masks localize each neuron, giving sharper edges, less background noise, and faster convergence on images, volumes, and SDFs.
Jun 1, 2026
A scale-space pipeline for terrain represented as Triangulated Irregular Networks (TINs): scale-aware seam-free sampling, an improved TIN smoothing operator, and a fully parallel GPU algorithm for tracking critical points across scales without global sorting.
May 8, 2026
