CV

Education

• BEng in Electrical and Electronic Engineering, University of Nottingham, Nottingham, United Kingdom, 2025–2027
• BEng in Electrical and Electronic Engineering, University of Nottingham Ningbo China, Ningbo, China, 2023–2027

Experience

• Research Assistant, Tsinghua University, Department of Automation, Beijing, China, Jul. 2026–Present
  • Supervisor: Prof. Keyou You
  • Participated in developing an end-to-end AI-driven control agent that automates controller design from non-expert system descriptions, covering problem diagnosis, controller selection, and closed-loop validation.
  • Established diagnosis and classification procedures to identify control problem types from system response characteristics, and designed safety-constrained, small-amplitude reversible experiments under limited input and state ranges.
  • Extracted core controller-relevant features from experimental responses instead of performing full system identification, and used these features for controller selection and tuning.
  • Implemented conservative controller synthesis, interactive closed-loop simulation, and performance evaluation, integrating the modules into a functional prototype spanning problem input, controller design, and closed-loop validation.
• Research Assistant (Remote), Tsinghua University, Department of Automation, Beijing, China, Sep. 2025–Feb. 2026
  • Supervisor: Prof. Keyou You
  • Built and debugged MuJoCo-based robotic manipulation environments for autonomous harvesting and flower-picking tasks in unstructured settings, using a Franka Panda manipulator for pick-and-place simulation.
  • Trained and compared Soft Actor-Critic (SAC) and Truncated Quantile Critics (TQC) policies, analysing training stability, robustness, and generalisation across simulated manipulation scenarios.
  • Identified catastrophic forgetting in SAC under selected training configurations and analysed policy stability and knowledge retention to better understand performance differences between reinforcement learning strategies.
  • Developed a custom 2-DOF robotic arm environment for sim-to-sim transfer exploration and analysed deployment bottlenecks caused by limited motion-capture support, including constraints on tracking accuracy and transfer evaluation.
  • Achieved a 100% task success rate with 25+ cumulative reward within approximately 1M training steps, validating the effectiveness of the established training and evaluation pipeline under the tested setting.
• Visiting Undergraduate Student, Shenzhen Research Institute of Big Data, Shenzhen, China, Jun. 2025–Aug. 2025
  • Supervisor: Assoc. Prof. Ruoyu Sun
  • Conducted a literature review on recent large language model post-training paradigms, focusing on synthetic continued pre-training, catastrophic forgetting during fine-tuning, and knowledge retention.
  • Studied synthetic data generation methods such as EntiGraph and examined the MoFo optimiser as an approach to mitigating catastrophic forgetting and improving knowledge retention during fine-tuning.
  • Conducted local deployment and inference experiments with open-source large language models such as GPT-OSS, gaining hands-on experience with model deployment, testing workflows, and environment configuration.
• Research Assistant (Onsite & Remote), Xi’an Jiaotong University, Bioinspired Engineering & Biomechanics Center, Xi’an, China, Jul. 2024–Jan. 2026
  • Supervisors: Prof. Feng Xu; Asst. Prof. Bin Li
  • Contributed to OsteoSight, a label-free virtual fluorescence staining and biophysics-anchored osteogenic fate inference system for conventional microscopy images.
  • Participated in developing and evaluating a contrastive-learning-based virtual fluorescence staining pipeline to reconstruct YAP, F-actin, and nuclei signals from label-free microscopy images, supporting model training, debugging, and performance evaluation.
  • Conducted experiments on image translation and generative models, systematically comparing virtual staining results with baselines such as CycleGAN in terms of morphology preservation, geometric consistency, subcellular structure reconstruction, and visual fidelity.
  • Co-authored the manuscript “OsteoSight: Label-Free Cell Osteogenic Fate Inference from Conventional Microscopy,” with the framework achieving a 90.69% F1-score for osteogenic fate inference across multiple microscopy modalities, donor age groups, and cell types. WSCON was warm-started using 156 pixel-aligned DIC–fluorescence pairs and subsequently adapted using 2,106 unpaired wide-field phase-contrast images.
  • Explored generative-model-based biomedical image preprocessing and enhancement methods, including Stable Diffusion and FLUX, and deployed an interactive demo on Hugging Face Spaces for image enhancement and domain adaptation experiments.
• Intern (Remote), University of Science and Technology of China, Hefei, China, May 2024–Jul. 2024
  • Supervisor: Prof. Wei Sun
  • Trained a ResNet-50 model for computer vision classification tasks, achieving over 80% accuracy through systematic parameter tuning and optimisation.
  • Implemented Deep Q-Network (DQN) methods for the CarRacing environment by converting the original continuous control task into a discrete action space, and trained an agent that achieved a peak mean reward of nearly 900.
  • Built a multimodal image retrieval workflow using Claude Sonnet 3, matching user text queries with model-generated image descriptions to retrieve and filter images according to user-defined requirements.

Research Output

• OsteoSight: Label-Free Cell Osteogenic Fate Inference from Conventional Microscopy
  • Co-author; manuscript under submission/review

Awards

• Scholarships, University of Nottingham Ningbo China
  • University Academic Excellence Scholarship Winner (Provost’s Scholarship), 2023–2024 and 2024–2025
  • Zhejiang Provincial Scholarship, 2024

Skills

• Languages & Tools: Python, C/C++, MATLAB/Simulink, CMake, Git, Linux/Shell, Docker
• AI Engineering: PyTorch, Computer Vision, LLM Integration (APIs/Prompting), Applied Reinforcement Learning, NumPy, Pandas, Matplotlib, Scikit-learn, XGBoost
• Embedded & Robotics: STM32, ROS, Raspberry Pi, Arduino, LTspice, FPGA (Verilog), PID Control, MuJoCo