About
Who am I
I’m currently working at Argonne National Laboratory as a computational scientist. Before joining KLA, I worked as an algorithm engineer at KLA (KLA-Tencor) and a postdoc at Argonne National Lab.
My research
I am interested in the following areas of research:
- Advanced optimization algorithms and learned priors for phase retrieval in coherent diffraction imaging,
particularly ptychography.
- Deep image prior [Publication]
- Text2Img diffusion model as a generative plug-and-play prior for image enhancement [Publication]
- Distributed training and parallel computing for scientific problems on high-performance computers
- Large-scale training of a crystallography structural prediction model [Publication]
- Distributed ptychographic reconstruction [Publication 1] [Publication 2]
- Automation of imaging and materials characterization experiments
- Bayesian adaptive sampling for X-ray absorption spectroscopy [Publication]
- Multimodal LLM agentic workflows and system design for X-ray imaging beamlines [Publication]
- Physical modeling of complex image formation processes and their reconstructions
- Multi-slice modeling for 3D imaging of thick objects beyond the depth-of-focus [Publication]
- Automatic differentiation (AD) for solving inverse problems in imaging science
- An AD package that reconstructs multiple imaging techniques and solve various parameters in the image formation model [Publication]
- Image processing, computer vision, and parallel computing for large-scale tomographic reconstruction [Publication]
My software packages
- Pty-Chi: PyTorch-based ptychographic reconstruction package supporting both analytical and automatic-differentiation engines [GitHub]
- EAA: agent harness for experiment automation
- Adorym (archived): automatic differentiation package for various imaging setups and reconstruction parameters [GitHub]
- Tomosaic: extending the field-of-view of tomography through image stitching [GitHub]
My hobbies
I am a amateur digital artist that create illustrations and fanarts of my favorite ACG characters! My most handy creation tool is iPad Pro + ProCreate. I also used to work on music videos (After Effects), composing and arrangement (LogicPro) in the Vocaloid/SynthV community. Other than that, I also know a bit of 3D modeling and animation with Blender. For a (incomplete) list of my past works, please refer to the Hobby Portfolio page.
What I’d write in this site
- Numerical computation and digital image processing in Python;
- Statistical analysis and visualization of research data;
- Use of Tensorflow/Autograd and their applications in solving optimization problems;
- Basic parallelization design based on MPI;
- Building and configuration of scientific-computation-oriented development environment;
- Some random Linux tips.
Contact me
关于我
目前，我在阿贡国家实验室（Argonne National Laboratory）担任计算研究员。此前，我曾在 KLA（KLA-Tencor）担任算法工程师，也曾在阿贡从事博士后研究。
我的研究
我的研究兴趣主要包括以下方向：
- 用于相干衍射成像相位恢复的先进优化算法与学习型先验，尤其是叠层衍射成像（ptychography）
- 面向科学问题的高性能计算，包括分布式训练与并行计算
- 成像与材料表征实验的自动化
- 复杂成像过程的物理建模与图像重构
- 通过多切片建模，实现厚度超过焦深的样品的三维成像 [论文]
- 利用自动微分（AD）求解成像科学中的逆问题
- 开发支持多种成像方式的自动微分软件包，同时重构图像并求解成像模型中的多种参数 [论文]
- 用于大规模层析重构的图像处理、计算机视觉与并行计算 [论文]
我的软件项目
- Pty-Chi：基于 PyTorch 的 ptychography 重构软件包，同时支持解析计算和自动微分两种引擎 [GitHub]
- EAA：面向实验自动化的智能体运行框架。
- Adorym（已归档）：基于自动微分的软件包，支持多种成像配置与重构参数的求解 [GitHub]
- Tomosaic：通过图像拼接扩展层析成像的视场 [GitHub]
我的爱好
业余时间，我喜欢画插画，尤其是喜欢的 ACG 角色的同人图，最顺手的创作工具是 iPad Pro 和 Procreate。此前，我也在 Vocaloid / SynthV 社区做过音乐视频（使用 After Effects），以及作曲和编曲（使用 Logic Pro）。此外，我还会用 Blender 做一点三维建模和动画。
业余作品集 页面收录了我的部分过往作品。
本站内容
- 基于 Python 的数值计算与数字图像处理；
- 研究数据的统计分析与可视化；
- TensorFlow / Autograd 的使用及其在求解优化问题中的应用；
- 基于 MPI 的基础并行程序设计；
- 面向科学计算的开发环境搭建与配置；
- 一些 Linux 使用技巧。