👋 About
I am Zian Zhou, a B.Eng. student in Software Engineering at Zhejiang University.
My interests lie in machine learning, multimodal foundation models, multi-agent reinforcement learning, quantitative ML, and trustworthy AI systems.
📰 News
- Mar 2026: Joined CICC as a Quantitative Strategy Intern in Machine Learning Research, focusing on multi-agent RL for option hedging and ML-driven CTA pipelines.
- Jan 2026: Started an independent winter research project on multimodal fusion and explainability in medical AI in preparation for MICCAI.
- Aug 2025: Joined Hangzhou Yuanzhoufang as an Algorithm Intern and contributed to multimodal large models for spinal disease, leading to one ICLR 2026 submission.
- Jul 2025: Completed the NUS Summer Research Program in Computer Science; my final project ranked #2 in its track and received A+.
- Mar 2025: Began undergraduate research at the State Key Lab of CAD&CG, Zhejiang University.
📚 Selected Publications
Full publication list on Google Scholar.
* denotes distributed contribution.
Click image to enlarge
SpineBench: A Clinically Salient, Level-Aware Benchmark Powered by the SpineMed-450k Corpus
A clinically grounded multimodal benchmark and data ecosystem for vertebral-level reasoning in spinal diagnosis.
Click image to enlarge
Ivy-Fake: A Unified Explainable Framework and Benchmark for Image and Video AIGC Detection
A clinically grounded multimodal benchmark and data ecosystem for vertebral-level reasoning in spinal diagnosis.
💼 Intern
China International Capital Corporation (CICC)
- Developed a multi-agent reinforcement learning framework for option delta hedging.
- Built an end-to-end machine learning CTA pipeline for rebar options, covering raw data cleaning, Alpha158-style feature engineering, factor screening and synthesis, model training, and an event-driven backtesting framework.
- Validated nonlinear alpha signals in RB options and improved strategy robustness through feature-importance analysis and systematic stress testing.
Hangzhou Yuanzhoufang Co., Ltd.
- Contributed to the development of a multimodal large model for spinal disease, leading to one ICLR 2026 paper submission.
- Constructed SpineMed-450k with 400K+ samples from textbooks, guidelines, and real-world cases using PaddleStructureV3 OCR and Gemini-2.5-Pro for traceable QA generation.
- Co-developed SpineBench for diagnosis and report generation, with an XML-based LLM-as-a-Judge evaluation protocol for reproducibility and objectivity.
- Explored agent memory through reproductions and literature study on Agentic Context Engineering and training-free GRPO.
Pi3Lab
- Independently built scholarwiki.com, including high-concurrency crawlers that ingest the latest arXiv papers and code daily.
- Automated paper summarization and social-media publishing workflows with n8n.
- Delivered enterprise-facing AI features for clients including China Telecom and Nike, covering AI design and automatic PPT generation modules.
🔬 Research
Independent Research for MICCAI Preparation
- Investigated multimodal fusion strategies for medical AI systems, focusing on aligning image, text, and structured clinical signals under limited-label settings.
- Reviewed and implemented representative designs for cross-modal interaction, feature aggregation, and uncertainty-aware interpretation.
- Studied explainability methods for multimodal medical models, including token-level attribution, region-level attribution, case-based analysis, and clinically meaningful rationale generation.
State Key Lab of CAD&CG, Zhejiang University
- Worked under Prof. Yingcai Wu on vision-language models for Chinese traditional painting, with a focus on fine-grained multimodal understanding.
- Participated in research under Prof. Shuiguang Deng on deep imbalanced regression and algorithmic improvements for long-tail data distributions.
🎓 Education
Zhejiang University
- Honors: National Scholarship, and Top 10 Students of Yunfeng College.
- Competitions: Gold Award in the Zhejiang Provincial International College Students’ Innovation Competition; First Prize in the National College Students Mathematics Competition (Zhejiang); First Prize in the Zhejiang Higher Mathematics Competition; Third Prize in the Zhejiang University Mathematical Modeling Competition; Third Prize in the Zhejiang Physics Competition.
National University of Singapore
- Final project ranked #2 in its track.
- Received A+, the highest individual grade.
🏆 Honors and Awards
- National Scholarship.
- Zhejiang University First-Class Scholarship.
- New Oriental Scholarship.
- Top 10 Students of Yunfeng College.
- Gold Award, Zhejiang Provincial International College Students’ Innovation Competition.
- First Prize, National College Students Mathematics Competition (Zhejiang).
- First Prize, Zhejiang Higher Mathematics Competition.
🛠️ Skills
- Programming: Python, C++, PyTorch, Selenium, Git.
- AI/ML: Multimodal LLMs/VLMs, reinforcement learning, quantitative ML, OCR pipelines, benchmark design, model fine-tuning, and backtesting systems.
- Research Interests: Multi-agent RL, machine learning, multimodal fusion, explainability, agent memory, and long-tail learning.
- Languages: Chinese (native) and English (strong academic reading and technical communication).