Wu Yanjing | National University of Singapore, Master of Economics

About Me

Data Scientist with over 2 years of experience in developing predictive models and analyzing complex datasets. I hold a Master’s degree in Applied Economics from the National University of Singapore (NUS), and I specialize in statistical modeling and machine learning. My background enables me to transform raw data into actionable insights, supporting strategic decision-making and driving business growth.

Selected Projects

Smart Property Insight: A Chatbot-Powered Real Estate Assistant

A GPT-4o-powered chatbot that helps users explore property listings and gain market insights through natural language queries. It automatically detects intent (listing vs analytics) and generates SQL to return relevant results. Example: “Find a condo in Orchard under $1M with a balcony” or “Which region has the most expensive properties?” Supports listing filtering, regional price analysis, and expandable features like LangChain-based memory and cloud deployment.

Code

Resume–Job Matching with Deep Learning

A semantic matching system that compares resumes and job descriptions using Sentence-BERT embeddings. It intelligently ranks candidates or job postings based on content relevance, helping recruiters and job seekers find better fits — beyond just keyword matching.

Code

ADHD Brain Connectivity Saliency with Deep Learning

Built a saliency-based analysis pipeline using an attention-based graph neural network to identify and visualize the most important brain regions distinguishing ADHD from control subjects. Leveraged fMRI connectome data from WiDS Datathon 2025 and visualized results using NetworkX and Nilearn. Evaluated model performance using F1 score and highlighted interpretable neural biomarkers from the trained model.

Code

HIV Epidemic Modeling with ART and Drug Resistance

A deterministic compartmental model inspired by Blower et al. (2000) simulating the long-term impact of antiretroviral therapy (ART) on HIV case counts and mortality. Includes ART initiation, drug resistance, dropout, and behavioral feedback. Simulations compare no ART, optimistic ART (low resistance, high adherence), and pessimistic ART (high resistance, risky behavior) using randomized parameters to reflect uncertainty. Results confirm ART's effectiveness only when paired with behavioral and adherence safeguards.

Code

Working Experience

Real Estate Analytics Pte. Ltd – Singapore

Data Scientist
Jan 2023 – Present

Data Analyst Intern – Data Science Team
Sep 2022 – Dec 2022

Machine Learning & Predictive Modelling
- Developed and optimized predictive models for housing demand forecasting, profitability analysis, and price index formulation using algorithms such as XGBoost, CatBoost, and Random Forest.
- Utilized NLP and large language models (LLMs) for advanced feature extraction, intent categorization, and other data processing tasks.
Big Data Engineering & Scalable Data Pipelines
- Designed and maintained ETL pipelines using SQL, Python, and Apache Airflow, ensuring seamless data ingestion, transformation, and storage for large-scale real estate data.
- Automated data pipelines with CI/CD integration using Git.
Data Visualization & Business Insights
- Created interactive dashboards and automated reports using Power BI and Tableau to present model outputs, performance trends, and strategic insights to both technical and non-technical stakeholders.