About Jiayi Xin
Hi, I’m Jiayi (Raina) Xin 👋
I’m a Ph.D. student in Computer & Information Science at the University of Pennsylvania, working with Prof. Qi Long. My research lives at the intersection of multimodal learning, large language models, and biomedicine—all with the goal of making AI that really helps patients and clinicians.
- I started my internship as a Machine Learning Intern at BigHat Biosciences! Exicited to apply my ML knowledge in solving challenges in antibody therapies.
- Interpretable Multimodal Interaction-aware Mixture-of-Experts (I$^2$MoE) got accepted into ICML 2025! I$^2$MoE achieves superior multimodal fusion performance and reveals how multimodal interactions influence predictions.
- Missing Modality Imputation with MoE-Retriever was awarded as the Best Paper at AAAI GenAI4Health Workshop 2025! MoE-Retriever fills in the gaps when some modualities, e.g., MRI scans and clinical notes, are missing.
- Estimate LLM Unseen Knowledge with KnowSum is now on arXiv! KnowSum estimates how much unseen knowledge today’s large language models know.
Academic Journey
- B.A.Sc. in Applied AI, The University of Hong Kong
- Capstone with Prof. Lequan Yu (HKU‑MedAI) & mentor Fuying Wang.
- Exchange & Research at Wellesley College and MIT
- AI for Science with Prof. Connor Coley, mentored by Samuel Goldman.
- Summer at the University of Oxford
- Clinical AI research mentored by Prof. David Clifton.
These experiences have shaped my interdisciplinary, people‑first approach to research and innovation.
Beyond the Lab
When I’m not debugging my models, you’ll find me running, bouldering, cooking, hunting for live music, or collecting passport stamps (I’ve travelled to the Philippines, Singapore, Malaysia, Thailand, Indonesia, Vietnam, Japan, Korea, Australia, and France). Each adventure keeps me curious—and reminds me why building AI applications that works for everyone matters.
Let’s Connect 🤝
I love swapping ideas, exploring new collaborations, and mentoring budding researchers.
- Email: jiayixin [at] seas [dot] upenn [dot] edu