Zixu Wang

I am a PhD candidate in Computer Science at Technical University of Munich and Infineon Technologies, where my work focuses on using large language models for automotive software development — from deciding what to build to specifying it. I am supervised by Prof. Dr. Chunyang Chen and expect to graduate in 2027. Alongside it, I serve as a teaching assistant for TUM’s course Foundations and Application of Generative AI.

Before the PhD I earned an MSc in Electrical and Computer Engineering at TUM and worked at Bosch Center for Artificial Intelligence, Siemens AG, and Infineon Technologies, where I developed trajectory-prediction and radar–camera depth-estimation models for autonomous driving, applied machine learning to logistics, and built AI models for sensor data. That path took me up the automotive stack — from how a vehicle perceives the world to how its software gets specified — and it shaped how I work today: on real engineering data, under real constraints.

Feel free to reach out for collaborations, questions, or just to say hi at hi@zixu.wang.

selected publications

ASE
Bridging Stakeholder and Product Requirements: An Empirical Study of Requirement Engineering in the Automotive Industry

Zixu Wang, Shengcheng Yu, Zhenchang Xing, Tobias Wenzel, and Chunyang Chen

In 41st IEEE/ACM International Conference on Automated Software Engineering (ASE), Industry Showcase, 2026

Abs arXiv Bib

An empirical study of how stakeholder-level requirements are evaluated, refined, and transformed into product-level requirements in industrial automotive practice, based on over 8,000 stakeholder and 5,800 product requirements with traceability links and decision outcomes.
@inproceedings{wang2026bridging, title = {Bridging Stakeholder and Product Requirements: An Empirical Study of Requirement Engineering in the Automotive Industry}, author = {Wang, Zixu and Yu, Shengcheng and Xing, Zhenchang and Wenzel, Tobias and Chen, Chunyang}, booktitle = {41st IEEE/ACM International Conference on Automated Software Engineering (ASE), Industry Showcase}, year = {2026}, }
arXiv
SocialFormer: Social Interaction Modeling with Edge-enhanced Heterogeneous Graph Transformers for Trajectory Prediction

Zixu Wang, Zhigang Sun, Juergen Luettin, and Lavdim Halilaj

arXiv preprint arXiv:2405.03809, 2024

Abs arXiv Bib

A graph transformer that captures semantic and spatial agent interactions for trajectory prediction in autonomous driving, combining temporal encoding and road topology to reach state-of-the-art results on the nuScenes benchmark.
@article{wang2024socialformer, title = {SocialFormer: Social Interaction Modeling with Edge-enhanced Heterogeneous Graph Transformers for Trajectory Prediction}, author = {Wang, Zixu and Sun, Zhigang and Luettin, Juergen and Halilaj, Lavdim}, journal = {arXiv preprint arXiv:2405.03809}, year = {2024}, }

RA-L

Semanticformer: Holistic and Semantic Traffic Scene Representation for Trajectory Prediction Using Knowledge Graphs

Zhigang Sun, Zixu Wang, Lavdim Halilaj, and Juergen Luettin

IEEE Robotics and Automation Letters, 2024

Bib

@article{sun2024semanticformer,
  title = {Semanticformer: Holistic and Semantic Traffic Scene Representation for Trajectory Prediction Using Knowledge Graphs},
  author = {Sun, Zhigang and Wang, Zixu and Halilaj, Lavdim and Luettin, Juergen},
  journal = {IEEE Robotics and Automation Letters},
  volume = {9},
  number = {9},
  pages = {7381--7388},
  year = {2024}
}