Yeqiu Wang
Sunnyvale, CA, USA | [email protected] | +1-401-328-1517 | linkedin.com/in/yeqiu-wang
Education
Brown University Providence, RI, USA
Ph.D. in Mathematics May 2025
The Chinese University of Hong Kong Hong Kong, China
B.Sc. in Mathematics, First Class Honours June 2019
Personal Projects
Large Language Model Q&A System for Corporate Banking Products May 2025 - June 2025
• Project Background: Corporate banking involves complex product structures and specialized domain knowledge. This project aimed to
develop a domain-adapted large language model (LLM) system to enhance client support through intelligent Q&A capabilities.
• Responsibilities:
– Fine-tuned the Baichuan2 model on 100,000 domain-specific Q&A pairs using LoRA within the LLaMA Factory framework
– Applied Direct Preference Optimization (DPO) with 10,000 financial consultation pairwise data to improve alignment with user
preferences
– Developed a Retrieval-Augmented Generation (RAG) system: encoded documents with the bge-large-zh embedding model and built
the vector database using FAISS, all orchestrated via LangChain
• Achievements: Boosted model performance from 67.2 to 87.8 on the TruLens evaluation framework, resulting in a robust and scalable
solution for complex banking product inquiries
Vision Permutator Implementation for Image Classification(GitHub Link) Jan 2025 - May 2025
• Developed a simplified Vision Permutator model with a team of four, replacing convolutions and attention with structured spatial and
channel permutations
• Trained models on MNIST (99.2% accuracy) and CIFAR-10 (78.5% accuracy), achieving competitive performance with fewer parameters
and FLOPs than CNN baselines
• Integrated advanced data augmentation techniques (CutOut, RandAugment, MixUp, CutMix) to enhance model generalization
• Optimized computational efficiency (0.15–0.45 GFLOPs) and benchmarked the Vision Permutator against CNN and Transformer baselines
• Presented project findings via a poster presentation
Gaussian Naive Bayes classifier(GitHub Link) Sep 2024 – Dec 2024
• Developed a Gaussian Naive Bayes classifier from scratch, achieving 100% test accuracy on wine dataset
• Collaborated with a team of four on model developing and evaluation using Scikit-learn benchmarks
Professiona Experiences
Lagrangian Plane Embedding in Hyperkähler Manifolds(PhD Thesis Research) Sep 2022 – Dec 2024
• Investigated Lagrangian embeddings of projective planes in the Hilbert scheme of K3 surfaces and their specializations
• Discovered new insights into hyperkähler manifolds, contributing to mathematical geometry research
• Analyzed complex geometric structures in the Hilbert scheme, processing 20,000+ examples in python
• Presented research through a poster at AGNES, a leading mathematics conference with 200+ participants
Fano Variety of Lines over C1 field Sep 2022 – Nov 2023
• Studied properties and structure of the Fano variety of lines over a C1 field
• Developed insights into rational points and their distributions
Teaching Assistant, Brown University Sep 2020 – May 2025
• Assisted in teaching three different calculus courses over six semesters, supporting classes of approximately 30 students each
Skills
• Statistics: Probability | Stochastic Calculus | Linear Regression | Polynomial Regression | PCA
• Machine Learning: AdaBoost | Neural Networks | Model Optimization | Random Forests | Transformers
• Programming: Python | Matlab | C | C++ | Octave | LaTeX | Mathematica | MS Office
• Python Packages: Numpy | Pandas | Scikit-Learn | Matplotlib | Tensorflow | PyTorch
• Language: Mandarin (Native) | English (Fluent) | Cantonese (Fluent) | French (Basic)
