Yijie Ding 丁一杰
pronounced as "yee-jee-eh ding"

I am currently an AI Engineer at Holotech on generative AI and extended reality. I am also completing my Master of Engineering at the College of Computing and Data Science at Nanyang Technological University, Singapore, supervised by Prof. Jianmin Zheng.

My research interests broadly lie in 3D computer graphics, with a focus on learning-based modeling and animation, including computer-aided design, motion generation, and interactive 3D systems.

I received my bachelor’s degree in Information Engineering and Media from the School of Electrical and Electronic Engineering, NTU. I was also a visiting student at ShanghaiTech University, where I had the honor of working with Prof. Xiaopei Liu and Dr. Kai Bai.

Email  /  GitHub  /  LinkedIn

Research

Multi-Agent CAD Code Generation

CAD-Factory treats CAD generation as a collaborative design process rather than a single-pass text-to-code task. It assigns specialized agents to plan model structure, instantiate geometric parameters, and validate executability, enabling complex CAD models to be built from simpler, coordinated subtasks with improved structural correctness and geometric consistency.

Bidirectional Query-Driven Generation of Parametric CAD Sketch

CADSketcher explores bidirectional CAD generation, treating any intermediate sketch state as a starting point for recovering both its possible history and future. Built on a query-driven framework, it completes partial parametric sketches with confidence-guided completion, executable instruction validation, and progressive context updates to improve geometric validity and sketch consistency.

ReACT: Reward-informed Autoregressive Decision CAD Transformer

ReACT brings decision making into CAD reconstruction: instead of imitating noisy traces, it learns to choose high-reward, intent-aligned commands that better match the underlying shape. This leads to cleaner and more faithful CAD programs, setting a new bar over prior imitation-learning approaches.

PartCAD: Learning CAD Modeling Sequences via Projection and Part Awareness

This paper presents PartCAD, a novel framework for reconstructing CAD modeling sequences from point clouds by projection-guided, part-aware geometry reasoning.

Academic Service

• Reviewer: CGI, SPM, PG, CVPR

Education

Nanyang Technological University (NTU) 2024 - Present
College of Computing and Data Science

• MEng (Computer Science)
• Thesis: Deep learning for intent-aware CAD reverse engineering

Nanyang Technological University (NTU) 2020 - 2024
School of Electrical and Electronic Engineering

• BEng (Information Engineering and Media), Highest Distinction
• Award: Dean's List (Academic Year 2022 - 2023)
• Award: Dean's List (Academic Year 2023 - 2024)

Projects

Selected projects from coursework and research.

Finger Soccer

A Unity-based soccer game controlled via Leap Motion Controller. Players use finger gestures to control characters—swinging fingers to walk and turn, mimicking kicks to shoot. The project demonstrates intuitive finger-based virtual embodiment and was showcased to Ubisoft during their visit to MICL lab.

GPU-Accelerated Volume Visualization on Multi-Resolution Grids

A GPU-accelerated volume renderer for OpenVDB data, implemented in CUDA and integrated with Aerocae. This project was conducted under the supervision of Prof. Xiaopei Liu and features adaptive sampling, customizable transfer functions, and isosurface rendering for scientific visualization applications.

Misc.

• I occasionally dabble in 3D modeling and animation with Blender. Feel free to explore my ArtStation.
• I am/was a vlogger on Bilibili I also do some photography in my free time. Some of my shots were featured by NTU (here and here). If you're on campus, swing by NTU's UShop for a Hall One postcard featuring my photo.

Template website from Jon Barron