Xuyang Chen

Munich, Germany · Open to opportunities

Xuyang Chen 陈旭阳

Generative AI Researcher & Engineer

Video DiffusionWorld Models3D / Neural RenderingLarge-Scale Training & Inference

Industrial Ph.D. researcher at TUM × Huawei Munich, combining publications with production-oriented experience in model training, adaptation, and deployment.

把人生当作马尔可夫链：只看当下，奔赴下一程。 “Treat life as a Markov chain — attend only to the present, and move on to the next state.”
Get in touch Google Scholar LinkedIn GitHub

01 About

My Journey

I was born in a small village in rural China. Driven by little more than my own motivation, I worked my way step by step — to school in the city, then to Harbin Institute of Technology, and on to Germany for graduate study, all the way to a Ph.D.

My background is unusually mixed: mechanical engineering at HIT, control and mechatronics at KIT, and now a Ph.D. centered on computer vision and generative AI at TUM × Huawei Munich. Each switch meant building the technical stack of a new field almost from scratch — and I have done it quickly every time. If there is one through-line in my path, it is adaptability: I learn fast, and I turn what I learn into working systems.

  • Ph.D. Generative AI & Computer Vision — TUM × Huawei Munich 2022–now
  • M.Sc. Mechatronics, Robotics & Automation — KIT 2018–2022
  • B.Eng. Welding Technology & Engineering — HIT 2013–2017

Research Focus

Perception, image / video diffusion for sim-to-real generation, and neural rendering / 3D Gaussian Splatting.

Engineering & Systems

Distributed training for generative models with PyTorch / Diffusers, DeepSpeed, Megatron-LM, and SLURM; inference deployment with TensorRT, vLLM serving, FP8 quantization, and custom CUDA operators.

02 Experience

Industrial Ph.D. Researcher

Jun 2022 – Present

Huawei Munich Research Center × TUM · Munich, Germany

  • Industrial Ph.D. researcher in Huawei Munich's generative AI / computer vision group, bridging research prototypes with production-oriented model training and deployment.
  • Research: perception; image / video diffusion for generation and sim-to-real adaptation; neural rendering and 3D Gaussian Splatting.
  • Engineering: distributed model training with PyTorch / Diffusers, DeepSpeed, Megatron-LM, SLURM; inference optimization with TensorRT, vLLM, FP8 quantization, custom CUDA operators, diffusion distillation, LoRA, and Dockerized environments.

Master's Thesis — Video Perception

Oct 2021 – Apr 2022

Porsche AG · Weissach, Germany

  • Video object detection with spatiotemporal Transformers; cross-frame attention for temporal coherence in automotive perception.

Research Intern — Radar Perception & Tracking

Nov 2020 – May 2021

Sony Electronics · Stuttgart, Germany

  • Deep-learning detection on sparse automotive radar point clouds; JIPDA multi-target tracking.

03 Publications

Published: 8  ·  bold = me · * = equal/co-first · corr. = corresponding author

04 Technical Skills

Training Systems

PyTorchDiffusersDeepSpeedMegatron-LMFSDP / DDPTP / PP / CP parallelismSLURMWeights & Biases

Inference & Deployment

TensorRTONNXvLLMCustom CUDA operatorsFP8 quantizationDiffusion distillationLoRADocker

Generative Models

Video diffusionImage diffusionRectified FlowStable DiffusionCogVideoXNVIDIA Cosmos-TransferFLUX.dev

3D / Vision & Languages

3D Gaussian SplattingNeural renderingOpen-vocabulary 3DObject detectionPythonC++CUDAEN / DE / 中文

05 Others

Let's build something.

Open to research & engineering roles in generative AI, 3D / vision, and large-scale model systems.

chenxuyang98@gmail.com LinkedIn Google Scholar GitHub