Jiayang Song

About Me

I am currently an Assistant Professor at the Leiden Institute of Advanced Computer Science (LIACS), Leiden University. My work connects Software Engineering and Artificial Intelligence to make AI systems safer, more reliable, and easier to understand.

Before that, I served as an Assistant Professor in the School of Computer Science and Engineering at Macau University of Science and Technology from 2025 to 2026. I received my Ph.D. from the University of Alberta in 2025, with a research focus on software engineering and intelligent systems, my M.Eng. from the University of Toronto in 2021, and my B.Eng. from Western University in 2019.

Research Interests

My research centers on Quality Assurance for Trustworthy AI Systems. I develop methods for testing, evaluation, analysis, repair, and enhancement to understand how AI systems behave and fail, and to improve their safety, reliability, robustness, and explainability throughout development and operation. Two questions guide my work: How can we establish trust in complex AI systems? How can we integrate AI reliably into the physical world?

My current work and future directions span three connected areas:

• Trustworthy foundation models. Testing and understanding language and multimodal models, including uncertainty estimation, confidence calibration, multilingual safety alignment, modality bias, and video and audio understanding. I am also interested in how models reason internally and how their representations shape their behavior.
• Embodied AI and cyber-physical systems. Evaluating and improving AI systems that interact with the physical world, particularly autonomous driving and robotic manipulation. This includes testing vision-language-action models, reconstructing simulation scenarios from real-world data, monitoring safety, and repairing unsafe behavior.
• AI for software and system development. Using foundation models to support task planning, reward design, and software development, while building quality assurance into these workflows. A future direction is test-driven human-AI collaboration, where testing provides evidence for assessing and improving AI-generated programs.

My long-term vision is to establish quality assurance across the full AI system lifecycle and enable AI agents to work reliably with people, software, and physical environments. This includes advancing models that can reason about physical laws and causal relationships, and understanding how trustworthiness emerges from the interactions among these components.

Prospective Students

I am looking for self-motivated PhD candidates with experience in software engineering and AI who are interested in advancing quality assurance for trustworthy AI systems. Master’s and undergraduate students are also welcome to contact me about research projects in these areas.

I also welcome inquiries from self-funded PhD candidates and those with external funding, such as a CSC scholarship.

Interested candidates please email me with your CV, transcript, a short (1–2 page) research proposal outlining your research interests, and sample publications (optional).