Experience

Industrial Experience

  • Applied Scientist, 2025 - present.
    Amazon, Rufus, CA, USA
    Responsibility:
    • Built the full coding RLVR recipe for our ~800B MoE model — verifiable-problem curation, verifier and reward design, and GSPO training at scale
    • Diagnosed and fixed reward-integrity failures at large scale (reward-grading starvation, truncation-driven response collapse)
    • Developed SFT training and data pipelines for large MoE reasoning models
    • Lead and execute in-house LLM CPT/mid-training and long-context extension recipes
  • Applied Scientist Intern, 2024 - 2025.
    Amazon, Rufus, CA, USA
    Responsibility:
    • Research SOTA linearizing and hybrid LLMs
    • Propose and implement a new hybrid LLM architecture
  • Research Intern, 2024.
    Visa Research, CA, USA
    Responsibility:
    • Develop a retrieval-augmented generation (RAG) framework to identify and rule out unrelated chunks without tuning
    • Explore potential application scenarios of RAG for business models
  • Backend Engineer, 2019.
    EZTABLE, Taipei, Taiwan
    Responsibility:
    • Cloud infrastructure design and maintenance
    • Utilized CI/CD tools for automatic testing and exporting reports

Education

  • Ph.D. in Computer Science, 2021 - 2025.
    Texas A&M University, TX, USA

  • M.S. in Structures and Materials, 2013 - 2015.
    National Cheng Kung University, Tainan, Taiwan

  • B.S. in Civil Engineering, 2008 - 2013.
    National Central University, Taoyuan, Taiwan

Research Experience

  • Graduate Research Assistant, 2021 - 2025.
    Texas A&M University, TX, USA
    Responsibility:
    • Developed model-agnostic algorithms that focus on generalization issues in machine learning models
    • Developed fairness machine learning frameworks for healthcare tasks by task-specific regularizations
    • Proposed the efficient pre-training paradigm for large language models during pre-processing
  • Research Assistant, 2019 - 2020.
    Institute of Information Science, Academia Sinica Topic: Conversational Information Retrieval Systems & Extreme Weather Prediction