cv

Basics

Name Ziye Chen
Email ziye2chen@gmail.com
Phone +1 (857) 961-9101
Url https://ziye2chen.github.io
Summary A Master's student in Artificial Intelligence at Boston University

Education

  • 2023.09 - 2025.01

    Boston, MA

    M.S.
    Boston University
    Artificial Intelligence
  • 2019.09 - 2023.06

    Nanjing, China

    B.S.
    Nanjing Tech University
    Mathematics

Publications

  • 2024.12.28
    Large Language Models for Mathematical Analysis
    arXiv
    In this research, we developed the DEMI-MathAnalysis dataset, comprising proof-based problems from mathematical analysis topics such as Sequences and Limits, Infinite Series, and Convex Functions. We also designed a guiding framework to rigorously enhance LLMs' ability to solve these problems.

Projects

  • 2024.09 - 2025.01
    Humanity’s Last Exam
    Contributed to a question-generation system for AI model evaluation
    • Developed a question-generation system for AI model evaluation
    • Contributed to the evaluation of AI models
  • 2024.01 - 2024.06
    Chinese OCR through a CRNN Architecture with Attention
    Implemented a CRNN with an attention mechanism for improved text recognition
    • Implemented a Convolutional Recurrent Neural Network (CRNN) architecture, enhancing the contextual understanding of text
    • Enhanced the model with an attention mechanism to improve focus on relevant segments of the input
    • Created schematics of the model structure
  • 2024.01 - 2024.06
    Election Narratives through Prompt Engineering with GPT
    Used GPT-based analysis on election-related emails, generating weekly insights on political parties
    • Used Gmail API to extract required emails from election-related accounts as data
    • Created charts to analyze basic information about political parties
    • Used prompt-based analysis for weekly merged emails
  • 2023.09 - 2023.12
    UBC Ovarian Cancer Subtype Classification and Outlier Detection
    Applied advanced image compression and EfficientNet for improved ovarian cancer subtype classification
    • Compressed images by Gem pooling (each image larger than 1GB in size)
    • Enhanced image compression techniques and trained an EfficientNet_b0 model to increase classification accuracy
    • Utilized boosting techniques, increasing accuracy from 65% to 85%

Awards

Skills

Programming Languages
Python (Proficient)
MATLAB (Proficient)
R (Proficient)
C (Proficient)
C++ (Basic)
Java (Basic)
SQL (Basic)
Machine Learning Frameworks
PyTorch
TensorFlow
Tools
Visual Studio Code
PyCharm
GitHub

Languages

Chinese
Native speaker
English
Fluent

Interests

Badminton
Soccer
Basketball
Cooking