Jinxi Xiang

Postdoctoral Fellow · School of Medicine

Jinxi "Derek" Xiang 项进喜

Portrait of Jinxi Xiang

About

Jinxi “Derek” Xiang is a Postdoctoral Fellow at Stanford University School of Medicine, working with Prof. Ruijiang Li on AI for precision oncology. He earned his Ph.D. from Tsinghua University in 2021 and was previously a Senior Researcher at Tencent AI Lab, where he led computational pathology projects deployed in clinical settings.

He develops multimodal foundation models that integrate histopathology, spatial transcriptomics, proteomics, and clinical text to reveal tumor biology and predict treatment outcomes. His research combines representation learning and generative modeling to learn from multiscale and multimodal data. His goal is to translate these advances into tools for understanding tumor heterogeneity, refining cancer classification, and guiding treatment across diverse patient populations.

Research at the intersection of AI, pathology, and spatial omics for precision oncology, linking patient tissue, foundation models, and clinicians.AIPathologySpatialOmicsPrecisionOncologyPatient TissueClinicianFoundation ModelVirtual TissueMultimodalFoundation ModelsBiomarkerDiscoveryClinical DecisionSupport

Research

01

Vision–language foundation models for oncology

Clinical decisions draw on images and text together. I develop foundation models that learn jointly from pathology images and clinical language at scale, then transfer with little or no labeled data to diagnosis, prognosis, and treatment response prediction.

02

Virtual tissue and the tumor microenvironment

Histology, spatial transcriptomics, and proteomics are different views of the same tissue. I build models that learn them jointly, a virtual tissue that can translate between modalities, generate missing ones from routine H&E, and map the cellular and molecular architecture of the tumor microenvironment at scale.

03

Generalizable learning from whole-slide images

Gigapixel slides with only slide-level labels demand learning methods that are data-efficient and robust across hospitals. I design multiple instance learning frameworks that turn foundation-model features into reliable, uncertainty-aware clinical predictions.

All publications →

News

  1. NIH K99/R00 application AI-Enabled Multimodal Spatial Biomarker Discovery for Precision Immunotherapy received an impact score of 18.
  2. Invited to serve as an Area Chair for ICML 2027.
  3. nnMIL: A generalizable multiple instance learning framework for computational pathology published in Nature Biomedical Engineering.
  4. Invited to serve as an Area Chair for ICLR 2027.
  5. A unified vision-language model for precision oncology and biomarker prediction in neuroblastoma (NEVA) published in Nature Communications.
  6. Recognized as a Gold Reviewer for ICML 2026.
  7. Two new preprints on multimodal histopathology AI: STORM and MUPAD.
  8. AI-enabled virtual spatial proteomics from histopathology for interpretable biomarker discovery in lung cancer (HEX) published in Nature Medicine.
  9. Received the IEEE Transactions on Medical Imaging Reviewer Certificate of Excellence 2025.
  10. Invited to serve as an Area Chair for ICLR 2026.
  11. Nature Cancer highlighted foundation models such as MUSK as substantial progress in digital pathology.
  12. MUSK featured by Stanford Medicine, Stanford Report, Stanford HAI, NVIDIA, The Stanford Daily, BioArt, and AdvanceBC.
  13. A vision–language foundation model for precision oncology (MUSK) published in Nature.

Invited Talks

  1. Northwestern University Chicago

    Decoding the Spatial Biology of Cancer Using Multimodal Foundation AI

    Hosted by Dr. Lee Cooper, Department of Pathology

  2. MD Anderson Cancer Center Houston

    Multimodal Foundation Models for Precision Oncology

    Hosted by Dr. Han Liang, Department of Bioinformatics and Computational Biology

  3. Houston Methodist Houston

    Integrating Pathology and Spatial Omics for Precision Immunotherapy

    Hosted by Dr. Keith Chen, Department of Urology

  4. AstraZeneca Palo Alto

    Multimodal Foundation Models for Precision Oncology

    Hosted by AstraZeneca Leadership & Stanford Medicine Industry Relations

  5. University of Warwick Virtual

    Unified Vision-Language Modeling for Computational Pathology

    Hosted by TIA Centre for Computational Pathology

  6. AbbVie Virtual

    Large-Scale Vision-Language Pretraining on Unpaired Pathology Data

    Hosted by AbbVie Monthly Imaging Seminar, Computer Vision Research Team

  7. Emory University Virtual

    Integrating Pathology and Clinical Notes for Outcome Prediction

    Hosted by Dr. Anant Madabhushi, Emory Empathetic AI for Health Institute

  8. Shanghai Jiao Tong University Virtual

    The Architecture of MUSK: A Multimodal Transformer Foundation Model

    Hosted by Dr. Weidi Xie, School of Artificial Intelligence

  9. Shanghai Jiao Tong University Virtual

    AI-Driven Molecular Biomarker and Relapse Prediction in Oncology

    Hosted by Dr. Zhou Chen, School of Biomedical Engineering

  10. University of Pennsylvania Virtual

    Predicting Immunotherapy Response via Multimodal Clinical AI

    Hosted by Dr. Zhi Huang, Perelman School of Medicine

Experience

Positions

  1. Stanford University

    Postdoctoral Fellow

    – Present · 2 yrs 11 mos

    Medical AI for precision oncology, with Prof. Ruijiang Li

  2. Tencent AI Lab

    Senior Researcher

    – · 2 yrs 7 mos

    Computational pathology, image/video coding, video generation

  3. SenseTime

    Research Intern

    – · 5 mos

    Medical image analysis

Education

  1. Tsinghua University

    Ph.D. in Engineering

    –

    Machine learning; computational imaging

  2. The University of Edinburgh

    Visiting Ph.D. Student

    –

    Electrical and Electronics Engineering

  3. University College Dublin

    Exchange Student

    –

    Electrical and Electronics Engineering

  4. Wuhan University

    B.E. in Engineering

    –