CV

Ph.D. Candidate

iCIS, Radboud University, Nijmegen, Netherlands

Master of Computer Science (Artificial Intelligence / Machine Learning)

University Paris-Saclay, Orsay, France

Bachelor of Opto-electronic Information Science and Engineering

Huazhong University of Science and Technology, Wuhan, China

Doctoral Researcher

iCIS, Radboud University, Nijmegen, Netherlands

• Developed FLOWSTEP, a contrastive learning and LLM-based framework that improves GitHub Actions workflow completion accuracy by 33.6\% over the state-of-the-art baseline.
• Developed CIGAR, a contrastive learning-based recommender system achieving 95.5\% top-5 accuracy in GitHub Action recommendation, outperforming GitHub Marketplace search by over 34×.
• Analyzed large-scale workflow execution data from 3,320+ repositories to study reruns, execution flakiness, wasted resources, and outcome prediction, achieving 90\% F1 score.
• Investigated 33K+ workflow security issues and recurrent weaknesses in CI/CD automation, supporting compliance-oriented analysis of software workflow practices.

Visiting Researcher

Hangzhou Dianzi University (HDU), Hangzhou, China

• Investigated AI techniques for automated workflow design and maintenance.
• Developed LLM-based methods for human-like test generation for open-source software.

Visiting Researcher

Nara Institute of Science and Technology (NAIST), Nara, Japan

• Designed FLOWSTEP, a step-type-aware framework for CI/CD workflow completion.
• Conducted large-scale empirical studies on the reliability of GitHub Actions workflows.

Research Scientist

Huawei Technologies France

• Designed reinforcement-learning-based scheduling algorithms for virtualized cloud infrastructure.

Research Intern

Group HUMANIA, INRIA/LISN

• Investigated privacy–utility trade-offs in traditional machine learning models.
• Designed and implemented an LTU attacker for membership inference attacks.