Xinge Yang (杨辛格)

I am a Member of Technical Staff at OpenAI. I received my Ph.D. from KAUST, where I worked with Prof. Wolfgang Heidrich in the Computational Imaging Group. My Ph.D. research focused on two directions: (1) AI for optics, using differentiable simulation and automated optimization to design optical systems, and (2) computational imaging hardware, building computational cameras by co-designing optics with neural networks.

My Ph.D. research advanced differentiable optics by establishing a complete simulation framework covering ray tracing, coherent and polarization tracing, sequential and non-sequential optical models, and ray–wave optics conversion. My representative work, published in Nature Communications, enables automated lens design via differentiable optimization. These efforts have been released as DeepLens, an open-source differentiable optical simulator integrating optics, sensors, and image processing for end-to-end imaging system design.

Email  /  Google Scholar  /  GitHub  /  LinkedIn  /  知乎

Xinge Yang profile photo

Work Experience

  • 08/2026 – present: Member of Technical Staff, OpenAI.
  • 07/2024 – 11/2024: Research Scientist Intern, XR Tech Camera & Sensing, Meta Reality Labs.
  • 10/2023 – 01/2024: Research Scientist Intern, Optics & Display Research, Meta Reality Labs Research.

Education

  • 2020 – 2026: M.Sc./Ph.D. in Computer Science, KAUST, Saudi Arabia.
  • 2016 – 2020: B.Sc. in Physics (major) and Computer Science (minor), USTC, China.

Research

My research develops AI methods for optics and uses them to build new imaging hardware:

  • AI for Optics: differentiable optical simulation, automated design algorithms, high-performance computing for optical simulation, and agentic workflows for optical design.
  • Computational Imaging Hardware: next-generation cameras built by co-designing optics and algorithms, including end-to-end simulation of optics, sensor, and ISP, high-fidelity synthetic data that bridges the sim-to-real gap, and efficient on-device algorithms.

Open Source
AI4Optics open-source ecosystem

AI4Optics
I summarize my PhD work on AI4Optics, with a mission to automate optical system design with AI agents and high-performance differentiable optics.

  • AI-native Optical Simulators: high-performance optical simulators built for LLM agents
  • Computational Imaging Applications: optical simulation combined with neural reconstruction.
  • AI Optical Design Software: AI-powered optical design in the browser with a built-in AI design assistant (DeepO).

Publications
Differentiable geometric waveguide display design result
Differentiable geometric waveguide display design

End-to-end differentiable design of geometric waveguide displays


Xinge Yang, Zhaocheng Liu, Zhaoyu Nie, Qingyuan Fan, Jim Bonar, Wolfgang Heidrich
arXiv preprint 2026. Paper (arXiv)
  • Differentiable Monte Carlo non-sequential polarization ray tracing for geometric waveguide display coating optimization with up to a thousand layers.
Task-driven lens design result
Task-driven lens design

Task-Driven Lens Design
Xinge Yang, Qiang Fu, Yunfeng Nie, Wolfgang Heidrich
Optics Express 2026. Paper (arXiv) / Paper (PDF) / Supp (PDF)
  • Task-driven lens design objective using a well-trained network to improve computer vision performance (image classification, object detection, and vision-language models).
Defocus deblur simulation result
Depth and spatially varying defocus deblur

Efficient Depth- and Spatially-Varying Image Simulation for Defocus Deblur


Xinge Yang, Chuong Nguyen, Wenbin Wang, Kaizhang Kang, Wolfgang Heidrich, Ginger Li
ICCV Workshop 2025. Paper (CVF) / Paper (PDF) / Supp (PDF)
  • Efficient and high-fidelity synthetic dataset generation for denoising and defocus deblurring without fine-tuning.
Aberration-aware depth from focus result
Aberration-aware depth from focus

Aberration-Aware Depth-from-Focus
Xinge Yang, Qiang Fu, Mohamed Elhoseiny, Wolfgang Heidrich
TPAMI 2025. Paper (IEEE) / Paper (PDF) / Supp (PDF) / Project page / Code
  • Efficient and high-fidelity synthetic dataset generation with neural PSF representation for depth estimation from defocus stacks.
Hybrid refractive-diffractive lens design result
Hybrid refractive-diffractive lens design

End-to-End Hybrid Refractive-Diffractive Lens Design with Differentiable Ray-Wave Model


Xinge Yang, Matheus Souza, Kunyi Wang, Praneeth Chakravarthula, Qiang Fu, Wolfgang Heidrich
SIGGRAPH Asia 2024. Paper (ACM) / Paper (PDF) / Supp (PDF)
  • Differentiable ray-wave optical model for hybrid refractive-diffractive lens design with two camera prototypes (extended depth of field and wide field of view).
DeepLens curriculum learning result
DeepLens ab initio deep learned refractive optics

Curriculum Learning for ab initio Deep Learned Refractive Optics
Xinge Yang, Qiang Fu, Wolfgang Heidrich
Nature Communications 2024. Paper (Nature) / Paper (PDF) / Supp (PDF) / Code / Video
  • Automated lens design from scratch with differentiable ray tracing and curriculum-based deep learning optimization.
Tolerance-aware deep optics

Tolerance-Aware Deep Optics
Jun Dai, Liqun Chen, Xinge Yang, Yuyao Hu, Jinwei Gu, Tianfan Xue
arXiv preprint 2025. Paper (arXiv) / Project page
  • Considering lens manufacturing tolerances in end-to-end lens design for better optical and computer vision robustness.
End-to-end optimization of fluidic lenses

End-to-end Optimization of Fluidic Lenses
Mulun Na, Hector Jimenez-Romero, Xinge Yang, Jonathan Klein, Dominik L. Michels, Wolfgang Heidrich
SIGGRAPH Asia 2024. Paper (PDF) / Project page
  • Differentiable optimization of liquid surface shapes and manufacturing for rapid prototyping.

Misc

  • Peer review statement: I am pleased to be invited to review papers in my research field. I will try my best to complete the review within two weeks. I apply the same high standards to every paper to maximize its potential. However, I do not expect the authors to address every suggestion I make, and I leave the final judgment to the associate editor.
  • Academic collaboration: I am open to collaboration and discussion, especially if my background in imaging and optics can benefit the project. I will try my best to put sufficient time into the project. If I simultaneously contribute to the idea, implementation, and paper writing, I expect to be listed as a co-first author; otherwise, I am comfortable with any authorship position.
  • More about me: I like cats, photography, new technology, and most sports (especially basketball and water sports). I have lived in China, Singapore, Saudi Arabia, the US, and Finland, and I want to live in different places to experience the lifestyle of local people. I enjoy adventures and challenges. I love my life.

Last updated: 10/14/2026 The website template is from Dr. Jon Barron.