Staff Engineer, Research · XRSpace

Kaiju Chen 陳楷儒

Bridges research and production — turning cutting-edge algorithms into deployed, scalable systems.

Engineer with 10+ years of experience in computer vision, 3D reconstruction, and deep learning. Delivered measurable impact for 40+ global fashion brands, reducing returns by 50% and increasing conversions by 30%.

Contact Me LinkedIn
13+
Years in Computer Vision & 3D
3
Research Publications
6
Engineers Led

Experience

Professional journey & key contributions

XRSPACE

Staff Engineer, Research

Jan 2024 – Present

Taipei City, Taiwan

Computer Vision Pose Estimation Motion Style Transfer Motion Generation Deep Learning Python PyTorch 2D Avatar
Style.me

Team Lead, 3D Scan & Reconstruction

July 2013 – December 2023

Taipei City, Taiwan

  • – Lead a team of six engineers developing 3D scanning and reconstruction algorithms, consistently delivering features on schedule.
  • – Architected and implemented a fully-automated clothing digitalization pipeline — 8× faster than manual post-production.
  • – Developed a fully-automated camera calibration algorithm (intrinsic & extrinsic) for RGB and depth cameras, saving 30 minutes of traditional setup per session.
  • – Built depth-camera-based reconstruction algorithms covering all clothing types.
  • – Generated 4K-resolution textures for 3D models using computer vision and deep learning pipelines.
  • – Trained deep learning models for alpha-matte prediction on translucent fabrics such as lace and chiffon.
  • – Implemented geometry processing algorithms: Laplacian smoothing, collision detection, collision handling, 3D object segmentation, and UV unwrapping.
  • – Developed GAN-based image inpainting to automate texture repair.
  • – Defined quality standards and built an automated quality-check scoring system for reconstruction results.
  • – Implemented face reconstruction: 3D avatar customization from a single front-facing image using 3DMM and SynergyNet.
  • – Developed tools for converting 3D models into AR-ready formats.
Computer Vision 3D Reconstruction Deep Learning GANs C/C++ Python TensorFlow PyTorch OpenCV
CyberLink

Software Engineer

Sep 2008 – June 2011

New Taipei City, Taiwan

  • – Implemented application-layer features for PowerDVD 8, 9, and 10.
  • – Built UI components including file browser, file/DVD/BD/UPnP playback using low-level Win32 APIs.
  • – Analyzed and optimized PowerDVD startup performance.
C++ Win32 MFC PowerDVD

Skills

Technical expertise across the stack

Programming Languages

C/C++ Python Matlab

Frameworks & Libraries

TensorFlow PyTorch OpenCV MFC Win32

Developer Tools & Infrastructure

AWS GCP Firebase Vercel Docker Git Jenkins CMake SVN

Domain Expertise

Computer Vision 3D Reconstruction Deep Learning GANs 3DMM Camera Calibration Alpha Matting UV Unwrapping AR Assets Pose Estimation

Education

Academic background

Sep 2011 – June 2013

Ph.D. Candidate in Computer Science

National Chiao Tung University

Hsinchu, Taiwan

Research interests: neural networks, pattern recognition, optimization theory, artificial intelligence

Sep 2005 – June 2007

M.S. in Multimedia Engineering

National Chiao Tung University

Hsinchu, Taiwan

Thesis: Simulated annealing for pattern recognition and seismic applications

Sep 2001 – June 2005

B.S. in Computer Science

National Chiao Tung University

Hsinchu, Taiwan

Publications

Peer-reviewed research contributions

1

“Hough transform neural network for pattern detection and seismic applications”

Kou-Yuan Huang, Kai-Ju Chen, Jiun-Der You, and An-Ching Tung

Neurocomputing (Special Issue) — Vol. 71, Oct. 2008, pp. 3264–3274

2

“Simulated annealing for pattern detection and seismic applications”

Kou-Yuan Huang and Kai-Ju Chen

Journal of Information Science and Engineering — Vol. 25, No. 3, May 2009, pp. 793–805

3

“Multilayer Perceptron for Prediction of 2006 World Cup Football Game”

Kou-Yuan Huang and Kai-Ju Chen

Advances in Artificial Neural Systems, Hindawi Publishing — Volume 2011, Article ID 374816