cv

My Curriculum Vitae.

Table of contents

Basics

Name Yifang Qin
Label PhD. Student
Email qinyifang (at) pku.edu.cn
Url https://yifang-qin.github.io/
Summary A Ph.D. student at Peking University working on generative models, AI for Science, LLM post-training with reinforcement learning, and recommender systems.

Work

  • 2026.10 - Present
    Intern, Post-training RL Team
    Moonshot AI (Kimi)
    • Develop agent capabilities for scientific coding through post-training, focusing on code understanding, generation, and execution.
    • Collaborate with AI for Science companies to improve model and agent capabilities for specialized industrial applications.
  • 2026.05 - 2026.10
    Top Talent Intern
    MiniMax
    Contributed throughout the M3.1 Flash iteration.
    • Built an RSI-oriented post-training loop (task synthesis → verifier → SFT/RL) to train RL experts for optimization tasks, improving performance on GSO and FrontierEng benchmarks.
    • Performance optimization: verifiable task synthesis from GitHub PRs, hack-resistant speedup reward design, and SFT/RL training for codebase performance optimization.
    • Engineering optimization: cross-disciplinary optimization task synthesis and simulator-based verifiers for SFT/RL training on simulation/design and AI4Science tasks.
  • 2025.10 - 2026.03
    Visiting Graduate Scholar
    UCLA Data Mining Lab, Los Angeles, US
    Supervised by Prof. Yizhou Sun.
    • Modeling reaction dynamics with flow-based models.
    • Reinforcement learning-based reaction trajectory generation.
  • 2024.05 - 2025.06
    Research Intern
    Microsoft Research AI for Science, Beijing, China
    • Science foundation models and machine learning for crystal structure prediction.
    • Energy-guided generation for chemical systems.
  • 2022.09 - 2023.12
    Research Intern
    Meituan, Beijing, China
    • Graph neural networks for Point-of-Interest recommendation.
    • Contributed to the online deployment of a billion-scale graph learning framework.
  • 2022.04 - Present
    Researcher
    Dlib Group, Peking University
    Supervised by Prof. Ming Zhang.
    • Graph-based recommender systems.
    • Deep generative models for natural science.

Education

  • 2023.09 - now
    PhD
    School of Computer Science, Peking University
    Computer Software and Theory
    • Supervisor: Prof. Ming Zhang
  • 2019.09 - 2023.07
    Bachelor
    School of EECS, Peking University
    Computer Science
    • Thesis: Learning Graph ODE for Continuous-time Sequential Recommendation

Awards

Languages

Chinese (mandarin)
Native speaker
English
Fluent

Interests

Research Interests
Generative models (Diffusion / Flow Matching)
AI4Science
LLM post-training & reinforcement learning
Recommender systems