Malisa · Design Engineering · 3D Computer Vision
I explore the intersection of 3D vision and design, building structured models that understand and generate spatial worlds.
I am pursuing an MSc in Design Engineering at Imperial College London after completing a bachelor's degree in Environmental Design at the School of Architecture, Tianjin University (GPA: 87.74/100). As a Research Assistant at Tsinghua University, I work on 3D mesh generation, 3D scene understanding, and Gaussian splatting. My research draws on design as a way of reasoning about geometry, structure, and how people inhabit space. I am seeking PhD opportunities to develop more structured and semantically grounded models of the 3D world.
My research examines how 3D representations can become more efficient, structured, and semantically grounded. Current work spans mesh generation, open-vocabulary scene reasoning, generative architectural modelling, and computational analysis of imaging data.
A category-specialized, reward-weighted fine-tuning framework for generating visually faithful meshes with lower structural complexity directly from images.
An object-centric Gaussian world model that couples attributed representations with open-vocabulary reasoning for compositional understanding of 3D scenes.
A generative framework for producing parametric 3D architectural models, connecting learned shape synthesis with editable design representations.
A computational approach to extracting neural signals and analysing functional connectivity patterns in calcium imaging data.
Selected projects in product, interaction, and spatial design, developed through research, prototyping, fabrication, and evaluation with users.
Ear-worn air purifier for second-hand smoke: full wearable industrial design plus a companion-app experience.
Figure-skating wearable + Apple Watch app, validated with on-ice user testing using real sensors and fabricated prototypes.
A narrative game world built in Rhino / SketchUp / Cinema 4D / Unity 3D — world-building, characters, and interactive scenes.
An RFID + sensor-driven installation that generates displaced cultural landscapes, exploring how places borrow one another's identity.
Additional work in interior, installation, and landscape design is available on request.
Education, research experience, publications, and selected design practice.
Download CV (PDF)