About Me

Hi, I am Jay Yixuan Duan.

I am a junior student at the University of Wisconsin-Madison. I am working with Professor Wei Qiu, Lei Hou, Hanwen Xu, and Sadeer Al-Kindi. Before that, I worked with Dr. Chao Wang. My work focuses on biomedical foundation models, interpretable AI, multi-omics integration, and AI for precision health. I am going to apply for 27 Fall PHD in Computer Science, Computational Biology and Bioinformatics.

🎓 Education

University of Wisconsin-Madison

• Expected to Graduate in May 2027, Bachelor of Science in Computer Science
• GPA: 3.88 / 4.0
• Relevant Coursework: Data Structures and Algorithms, Deep Learning, Artificial Intelligence, Big Data Systems, Human Computer Interaction, Advanced Natural Language Processing, Molecules to Life and Science

🚀 Research Interests

• Biological Foundation Model
• AI for Precision Health
• Interpretable AI
• Multi-omics Integration

📖 Publications

2026

CADENCE: A Cardiac Atom Dictionary for Interpretable Neural Concept Extraction from ECG Foundation Models

Yixuan Duan, Arjun Naik, Sadeer Al-Kindi, Wei Qiu

Under review. [arXiv]

• Built a dictionary of interpretable “cardiac atoms” to decompose the internal representations of ECG foundation models.
• Extracted human-understandable neural concepts from learned features and linked them to clinically meaningful cardiac patterns.
• Enabled interpretable analysis and concept-level probing of ECG foundation model representations.

2026

ECG-InterpBench: Benchmarking the Interpretability of ECG Foundation Models with Matched-Scale Sparse Autoencoders

Yixuan Duan, Wei Qiu

Under review. [arXiv]

• Proposed a capacity-controlled benchmark that compares representation-level interpretability across six frozen ECG foundation models under a matched sparse-autoencoder protocol.
• Built a 450-cell matched-scale interpretability atlas over five encoder depths, five dictionary widths, and three seeds, with leakage-controlled metrics for sparse reconstruction fidelity, single-feature clinical concept accessibility and coverage (49 ECG measurements), and cross-seed reproducibility.
• Provided robust comparisons using patient- and design-level uncertainty, sparsity sensitivity, and an external MIMIC-IV-ECG replication.

💻 Research Experience

Rice University
Feb. 2026 - Present, Worked with Professor Wei Qiu
ECG Foundation Models · Sparse Autoencoders · Interpretability

Weill Cornell Medicine & Houston Methodist Hospital
May 2026 - Present, Worked with Sadeer Al-Kindi
Echocardiography · ECG · Interpretability · Foundation Models

Boston University
Jan. 2026 - Present, Worked with Professor Lei Hou
Multi-omics Integration · scATAC-seq · DNA Methylation

Shenzhen Bay Laboratory
Jun. 2025 - Aug. 2025, Worked with Dr. Chao Wang
Glycan–Protein Binding · Molecular Modeling

💪 Skills

• Linux, Python, R, Java, SQL
• ECG and PPG analysis, RNA-seq, scRNA-seq, scATAC-seq, DNA methylation

🏆 Awards

• 2026 UW-Madison Dean’s List
• 2026 UW-Madison Undergraduate Summer Award