---

About Me

I am currently a second-year undergraduate at UCL in the MEng Mathematical Computation (new name as Computer Science and Mathematics) programme. My major interests lie in the field of mathematical foundations of computing, machine learning and educational system design. I enjoy solving problems and turning ideas into both proofs and codes. In my free time, I also like to study languages, explore societal topics and read books. I am currently investigating the role of institutional frameworks in governing and securing AI agent networks. Download My CV | Download My CV (Chinese)

Research Interests

Research Interests I am preparing for my long-term research interests in the mathematical foundations of machine learning, with a particular focus on model verification, safety, and the ethical implications of algorithms. I am also interested in applying game-theoretic modelling to learning systems, especially in the context of educational fairness and system design.

Publications

• Pu, H. & Yi, K. (2024). “A Comparative Analysis of EfficientNet and MobileNet Models’ Performance on Limited Datasets: An Example of American Sign Language Alphabet Detection.” Highlights in Science, Engineering and Technology, 94, 558–564. DOI: 10.54097/yh5d3s04
  • Research Content: This paper compares the performance of two major light CNN architectures, EfficientNet and MobileNet, under the constraint of limited datasets. This experiment has used several datasets downloaded and merged from Kaggle, and has trained the models under 5, 10 and 20 epochs. The result has shown that MobileNet V2 outperforms other all other models in testing accuracy under 5 epochs, but all other models suffer from severe overfitting problem after 10 epochs.
  • Research Impact (as of Nov 30th, 2025): 8 citations recorded on Google Scholar, 496 publisher-page downloads within the first year of publication.

Research Projects

• Systematic Review of Numerical Information Extraction from Biomedical Literature - 2026 (Ongoing)

  • Participating in a systematic review examining how natural language processing (NLP) methods extract numerical information from biomedical literature, including drug dosages, concentrations, laboratory values, and tumour measurements.
  • The review evaluates the robustness, safety, and sustainability of these methods, with a focus on data provenance, annotation quality, evaluation validity, and reproducibility.
  • Working under the guidance of PhD researcher Lara Carter from UCL, contributing to the systematic review and manuscript preparation as a coauthor.

• American Sign Language Spelling Recognition System + Game - 2024

  • Final project of the research project "Introduction to Machine Learning and Neural Networks", led by Professor Rogers from University of Oxford.
  • This project has used a low-code machine-learning platform, EdgeImpulse, to build a machine-learning algorithm based on MobileNet V2 96x96 0.25 that can recognise the gesture of the basic alphabet gestures of American Sign Language. We have then deployed the model on a web application and created a game to help users learn and practice the gestures.
  • Achievement: Highest Grade Awarded Among All Groups in the Project
  • My Role: Model Training (with other Teammates), Frontend Construction & Model Deployment, Final Viva Presentation.

• MVITA - Multi-Cultural Virtual Interview Agent, Hackathon Project - 2025

  • A project that helps the jobseekers who want to work overseas to prepare for the interview and adapting to other country's working culture through practising cultural-based interview scenario.
  • We have used Vue.js + Anam SDK as the frontend to achieve the user interface and the model's interaction with the user, as well as some logical implementations. We have also used Python as the backend to connect to ChatGPT's API to use GPT-5 model to generate interview questions based on cultural context and the user's CV. We have also used Flask as the backend to create a RESTful API to connect the frontend with the backend.
  • My Role: Frontend Construction & Anam SDK Connection. Prompt Engineering of the Question-Asking model and the Interviewer-Model.

Education

• University College London, United Kingdom - MEng Mathematical Computation (2025-2029)
  • Currently: Year 2 Student
  • Key Modules Taken (or Yet to be Taken This Year): Security, Logic, Computer Architecture and Concurrency, Theory of Computation; Algebra 3: Further Linear Algebra, Probability and Statistics, Analysis 1
  • Core Modules This Year
    • Mathematics: Algebra 3: Further Linear Algebra, Computational Methods and Programming, Probability and Statistics, Mathematical Methods 2
    • Computer Science: Logic, Security, Directed Reading, Computer Architecture and Concurrency
• Dulwich College (Singapore), Singapore - IGCSE & IB Diploma (2021-2025).
  • IBDP Bilingual Diploma, 41 Points
  • IB Subjects, Marks Awarded:
    • Higher Level: Mathematics: Analysis & Approaches, 7; Computer Science (with Options D - OOP), 7; Physics, 6;
    • Standard Level: Business Studies, 7; Chinese A: Language & Literature, 6; English A: Language and Literature, 6;

Skills

• Programming: Python, Vue.js, JavaScript, TypeScript, Node.js, Express.js, Java, C, Haskell
• Languages: Chinese (Native), English (C2), Japanese (B1~B2)

Roles, Teachings and Outreach

• UCLCSSA (UCL Chinese Students and Scholar Association), Department of Mandarin Education, Head of Department @ UCLCSSA, 2026 - now
  • Our aim is to foster China-UK cultural exchange in UCL. We are organising and coordinating activities such as the Golden Autumn Fair collaboratively with the Department of Culture of UCLCSSA, which will be held in November at the Bloomsbury campus. Besides that, we are also discussing seminar-style activities with the Department of Academia of UCLCSSA, whose aim is to allow the UCL community to understand China from a Chinese angle in greater depth. We also monitor the teachings on a regular basis.
• UCLCSSA (UCL Chinese Students and Scholar Association), Department of Mandarin Education, Higher Level (高级班) Teacher @ UCLCSSA, 2025 - 2026
  • In this course, I teach some advanced topics about Chinese Language Acquisition. In the lessons I am in charge of, I started off with some higher-level Chinese grammar and vocabulary, encouraging them to read longer, real-life passages in Chinese, along with some interactive activities encouraging them to learn about China and the Chinese culture. Then, we moved to Chinese Language and Literature, with the focus being on the book "Taipei People" (《台北人》). We explored reading and analysing Chinese literature texts and more about Chinese culture in its historical background, including the shared roots and cultural ties across the Taiwan Strait.

• Dulwich College (Singapore) - HiMCM Coordinator and Mathematical Modelling Teaching Lead - 2024
  • In this course, I am in charge of coordinating the participation of HiMCM (The High School Mathematical Contests for Modelling) in DCSG. I have taught several techniques required for HiMCM such as AHP and TOPSIS, trying to stimulate their interest in Mathematical Modelling and Applied Mathematics, incorporating the teaching in understanding the underlying principles of how those models work, instead of simply teaching the models and let the students memorise.