Hello, I'm Shiyang. 👋
I am a Ph.D. student in Professor David van Dijk's lab at Yale University. My work lies at the intersection of Large Language Models, Deep Learning, and Agent Systems.
Prior to Yale, I earned my M.S. from Columbia University and hold a B.Sc. in Statistics from the University of British Columbia. My research focuses on developing foundation models and agentic systems to unravel the complexities of biological systems.
🔥 News
2026.08
Contributed to the development and release of Apodex 1.1 during my summer research internship at Apodex. 🚀
2024.09
Started Ph.D. journey at Yale University in Prof. David van Dijk's lab! 🎉
2024.05
Graduated with M.S. from Columbia University. 🎓
💼 Experience
Apodex
Research Intern
- Contributed to the development and release of Apodex 1.1: Scaling Agentic Intelligence for Complex Work, supporting the end-to-end model development process across large-scale data construction, mid-training, post-training, and agent-system evaluation.
- Built and iterated on 50B-token-scale data synthesis pipelines for mid-training and post-training, spanning data generation, filtering, quality control, task construction, and dataset validation across tool-use, health & clinical, and agentic reasoning domains.
- Developed large-scale tool-use training data and evaluation workflows, constructing diverse tool-interaction trajectories and multi-step tasks to improve models' ability to select, invoke, and coordinate tools in complex execution environments.
- Contributed to health & clinical data synthesis for model training through domain-specific data construction, quality filtering, and evaluation, while supporting broader AI4Science projects with scientific data generation and model evaluation workflows.
- Implemented and improved the agentic system and execution harness for complex, long-horizon work, including task decomposition, tool interaction, and intermediate-state management.
- Designed evaluation and iteration pipelines to systematically analyze agent failures, compare model checkpoints and data recipes, and identify bottlenecks in reasoning, tool selection, and end-to-end task completion.
📖 Education
Yale University
2024 - PresentPh.D. in Computer Science
Advisor: Prof. David van Dijk
Columbia University
2023 - 2024M.S. in Operation Research
University of British Columbia
2018 - 2023B.Sc. in Statistics
🌟 Selected Publications
Apodex 1.1: Scaling Agentic Intelligence for Complex Work
arXiv preprint, 2026
A general-purpose model and execution system for long-horizon, verifiable work. My contributions covered large-scale data construction, model training, agent-system development, and evaluation.
Intelligence at the Edge of Chaos
arXiv preprint, 2024
A novel framework exploring the intersection of complexity theory and neural network dynamics.
Scaling Large Language Models for Next-Generation Single-Cell Analysis
bioRxiv, 2025
Developing foundation models specifically designed for single-cell genomics data analysis.
RankAdaptor: Hierarchical Dynamic Low-Rank Adaptation for Structural Pruned LLMs
arXiv preprint, 2024
Novel approach for efficient fine-tuning of large language models through dynamic rank adaptation.
📝 All Publications
Loading publications from Google Scholar...