Work and Service

Work Experience

AI Research Internship, Data & Data Science Group, Sanofi,S.A., 06/2023 ~ 09/2024, Boston, MA

Mentor: Wei Zhao and Bozhao Qi

  • Endoscopic Scoring and Localization in Unconstrained Clinical Trial Videos (Accepted at WACV2025)
    • Collected a comprehensive clinical endoscopy dataset through human-in-the-loop active learning, combining video cleaning and weak labeling by doctors to produce a large and balanced dataset (~49.9 hours, ~86,423 clips) for endoscopic mayo scoring
    • Developed an end-to-end foundational model for endoscopic scoring and localization, incorporating a ViT-based scoring model for clip-by-clip assessment and a trajectory model for precise localization, also serving as a vision input for medical multimodal models
    • Introduced a novel personalized Cumulative Disease Score (CDS) based on movement, surpassing the limitations of prior time-based CDS
  • Data-driven Neural ODE (DN-ODE) Modeling for Clinical Trails Breast Cancer Tumor Dynamics (Accepted at Neurips DLDE, CBM)
    • Developed a breast cancer pharmacodynamics model using neural ODEs, trained on phase 1/2 Amcenestrant clinical trials and validated in phase 3, which predicts tumor size and progression-free survival (PFS) from initial patient status
    • Offered interpretable results and visualizations of tumor dynamics, serving as a robust reference for clinical trial design by identifying potential responders and patients at risk of cancer progression
  • Generative AI & Retrieval-Augmented Generation (RAG) web-based Backend
    • Built a chat-based tool for researchers and doctors to review clinical trial documents and interactively find related answers.
    • Developed AWS Lambda backends integrating OpenAI’s ChatGPT API for query embeddings and Pinecone for similarity search.
    • Implemented automated tests with SonarCloud checks, ensuring high-quality code through GitHub Actions CI/CD pipelines.
  • Data-Driven Neural-ODE Modeling for Breast Cancer Tumor Dynamics and Progression-Free Probability Predictions
    • Managed sparse data from early phases of clinical trials and predicted later phases with only a single observation
    • Created visualizations depicting various tumor dynamics patterns and achieved accurate predictions (R2 exceeding 0.9)
    • Provided a robust reference for clinical trial design by identifying potential responders and progressed patients
  • Unpaired Tissue Segmentation & Virtual Restaining
    • Proposed an end-to-end unpaired segmentation GAN and virtual staining framework that mimics real staining processes
    • Supervised prior works and presented a promising substitute for traditional histopathology methods for clinical trial recruitment

Machine Learning Engineer Internship, Kernel Lab, 01/2021 ～ 08/2021, Seattle, WA.

Mentor: Dennis Meng, Lead Data Scientist at Kernel Labs Inc

  • Gender classifier: Design a model to classify a speaker’s gender.
  • Echo challenge: Design a Dual-signal Transformation LSTM Network to solve the acoustic echo cancellation (AEC) problems.

Teaching Experience

Teaching Assistant, University of Washington, 01/2020 ～ Now, Seattle, WA.

  • EE 241 Python Programming
  • EEP 596 Recommender Systems, Instructor: Karthik Mohan
  • ME/EE 547 Linear Control Theory, Instructor: Xu Chen

Services

  • Conference Reviews: CIKM, KDD, NeurIPS, AAAI, MICCAI, CLAI, ICLR, ICML, CVPR, WACV, ICMD, ISKE, SAM
  • Workshop: NeurIPS AFT, NeurIPS AID4, NeurIPS DLDE, NeurIPS AI4D
  • Journal Reviews: PLOS ONE, Data Science and Management (DSM), BICM, TNNLS
  • Other Services: Graduate Applicant Support Program (GASP)