CV
Education
- M.S. in Johns Hopkins University, 2025
- B.S. in Xi’an Jiaotong-Liverpool University, 2023
Research Experience
- Sep 2024 – Present: Research Assistant
- Johns Hopkins University
- Research Topic: Polynomial optimization via the Product-of-Measures (POM) framework
- Duties included:
- Advanced the theoretical development of the POM framework as a scalable alternative to traditional SOS/moment hierarchies, targeting high-degree and large-scale polynomial optimization problems
- Investigated structural connections between POM relaxations, SOS/moment hierarchies, and algebraic geometry certificates, with a focus on rank-1 moment vectors and Dirac measure representations
- Designed algorithms to decompose global SDPs into local subproblems using tensor-product measures, enabling more efficient computation without sacrificing global optimality guarantees
- Conducted numerical experiments on benchmark and non-SOS polynomials (e.g., Motzkin polynomial), demonstrating both feasibility and interpretability improvements over classical approaches
- Explored the role of algebraic certificates (e.g., Markov–Lukács forms, secant varieties) in certifying low-rank solutions, bridging optimization with tools from algebraic geometry
- Supervisor: Mateo Díaz
- Spring 2024: Research Project – Music Emotion Prediction using Recurrent Neural Networks
- Johns Hopkins University
- Duties included:
- Designed and implemented deep learning models (RNN, BRNN, and LSTM) to classify music into four emotional quadrants based on Russell’s Emotion Model
- Extracted and engineered 14 key audio features (e.g., chroma, MFCC, spectral features) using Librosa, preparing datasets of 900 original clips, 3,600 augmented clips, and a large-scale 14,000-clip subset from MTG-Jamendo
- Conducted comparative analysis of neural network architectures against 12 baseline machine-learning classifiers, highlighting strengths and limitations of RNN-based approaches
- Applied data augmentation techniques (noise injection, pitch/time shifting, speed variation) to expand datasets, leading to a 20–30% improvement in model accuracy
- Evaluated scalability of models across small, medium, and large datasets, identifying trade-offs between simplicity (RNN) and complexity (LSTM/BRNN) for generalization performance
- Project Paper (arXiv:2405.06747)
- Jun 2022 to Sep 2022: Research Assistant
- Xi’an Jiaotong-Liverpool University
- Research Topic: Personal Thermal Comfort During Exercise
- Duties included:
- Designed a research protocol to measure participants’ thermal comfort during cardiovascular exercise
- Created regression models to assess the relationship between thermal comfort and heart rate
- Utilized multilayer perceptron models to process data, revealing that heart rate partially reflects activity intensity
- Supervisor: Long Huang
Work Experience
- Mar 2021 – Jul 2021: Software Development Intern
- Shanghai Connext Information Technology Co., Ltd., Shanghai, China
- Project I: Asics Data Visualization System
- Spearheaded the design and development of a company-wide data visualization platform for Asics, transforming raw transactional data into accessible business insights
- Engineered a one-click reporting pipeline by extracting and integrating order data from internal databases, significantly improving operational efficiency and enabling data-driven decision-making across teams
- Project II: Tmall Competitive Product Monitoring System
- Designed and deployed an automated system that delivered daily competitive intelligence reports via email to senior management, ensuring timely and actionable market awareness
- Enhanced monitoring capabilities by incorporating product image segmentation, merging, and store-level data acquisition, creating a more comprehensive competitor analysis pipeline
- Established a robust 24/7 monitoring framework capable of generating automated daily reports, improving leadership’s ability to track and respond to dynamic e-commerce competition
- Sep 2020 – Jan 2021: Data Mining Intern
- Shanghai Gildata Inc., Shanghai, China
- Project I: Universal Crawler Platform
- Developed a scalable and modular web crawler platform capable of extracting structured data from diverse Internet sources, including tax bureaus, banks, and brokerage websites
- Optimized the crawling process to handle heterogeneous data formats and sources, improving data reliability and scalability for downstream applications
- Deployed the platform in production to continuously collect financial and regulatory data, while also identifying and reporting system bottlenecks, thereby contributing to long-term platform stability and usability
Skills
- Computer Languages: Python, SQL
- Languages: Mandarin Chinese (Native), English (Proficient)
- Interests: Piano, Powerlifting (Asian Record Holder)
