CV
Research interests
Formal verification of neural networks, including Transformer verification, robustness analysis, and the specification of verification properties.
Education
Kyushu University
Ph.D. student, Oct 2024 – Sep 2027 (expected) · Fukuoka, Japan AI safety, formal verification of Transformers, and certified robustness.
Waseda University
Master’s degree, Oct 2022 – Sep 2024 Low-resource German–Upper Sorbian neural machine translation and case-based reasoning for neural machine translation.
Sichuan University
Bachelor’s degree in Cyberspace Security, Sep 2019 – Jun 2023 Wu Yuzhang Honors College; machine learning and fake-review detection.
Experience
AI Research and Engineering Intern · Corpy & Co.
Oct 2025 – Jul 2026 · Remote / Japan
- Developed and evaluated LLM- and OCR-based pipelines for structured document understanding.
- Compared language-model platforms and investigated robustness issues in structured extraction.
- Built data-generation workflows for PDF, Markdown, and spreadsheet sources.
- Contributed to retrieval-augmented question-answering workflows for technical documents.
Selected outputs
- CAV 2026 publication — Computer Aided Verification (CAV), 2026. A formal verification method for certifying the robustness of Transformer models through tighter bounds for attention-related nonlinear operations.
- SurveyFlow — An open-source local graphical workspace for building auditable literature surveys through multi-source discovery, filtering, human review, citation snowballing, and export.
Talks
- CAV 2026 Presentation — Computer Aided Verification, Lisbon, Portugal, Jul 2026.
- SAIV 2026 Short Presentation and Poster — SAIV 2026, Lisbon, Portugal, Jul 2026.
Award & funding
- Selected for the K-BOOST doctoral support program, 2025–2027.
Skills
- Programming: Python, Git, Linux
- Machine learning: PyTorch, Transformer models, natural language processing, large language models
- Verification & optimization: Neural network verification, bound propagation, linear programming, Gurobi, abstract interpretation
- Document AI & engineering: OCR, LLM APIs, vector databases, Docker
Languages
- Chinese: Native
- English: Professional working proficiency
- Japanese: JLPT N2
- Korean: Beginner