Welcome to my portfolio website showcasing my selected projects in AI/ML, research, and design.
I’m Becky Xu, currently at the World Bank Group’s applied AI lab (DIME), building ZeroHungerAI. Since 2022, I’ve focused on AI/ML + product for climate and sustainability, with additional experience in sustainable urban planning, renewable energy, decarbonization strategy, and net-zero design.
Happy to chat if you'd like to connect!
This project explores how strategic masking—simulating cloud covers and other occlusions—can enhance the robustness of a pretrained Satellite Machine Learning (SatML) model for land cover classification tasks through fine-tuning.
I explored the impact of physical and environmental factors by engineering features from the Boston Housing dataset and comparing the performance (RMSE) of LASSO Regression, Random Forest, and Ensemble models.
Carbon Offset Validator is a full-stack tool for analyzing forestry carbon projects by project code.
This platform visualizes risk metrics, land-use and deforestation trends, geospatial boundaries, and AI-generated insights.
View the site here!
Built for client UNOPS, this web tool powers project insight dashboard, covering ~9000 projects. A Next.js frontend visualizes project portfolios overview, SDG alignment, funding, regions, and tool usage, while a FastAPI/Python backend pulls and transforms BigQuery data. Used LLM with few-shot prompting for document-analysis and keyword extraction.
Curated over a dozen global datasets on climate risks and socio-economic conditions during my work at a climate-tech startup, AlphaGeo. I conducted research, data sourcing, and ETL to create 1km resolution datasets and automated pipelines. In addition, I developed the "resilience-adjusted risk score" methodology with the CTO to evaluate risks under climate change scenarios.
This project, developed for the MIT Civic Data & Design Lab (CDDL) in spring 2023, investigates the transformation of Beirut's urban form and its impact on small businesses and street commerce. I conducted spatial and data analysis, data visualization, and GIS mapping to analyze the spatial distribution of small businesses and street commerce in Beirut.
This project is developed for 11.220 Quantitative Reasoning class at MIT in fall 2022. I investigated the health insurance inequality in NYC using data from the American Community Survey (ACS) 2015 - 2019.
The Philly Energy Relief Hub is a one-stop shop for low income households to discover public benefit programs in Philadelphia to reduce the share of their income spent on energy, both in the short-term and long-run, and ease the process of determining eligibility and applying for programs that meet their needs. Developed this for MIT 11.138. I collaborated with three other students, took charge of front-end development, web design, and project management.
Redesigned the MIT Energy & Climate Club website to improve user experience and visual design.
During NYC Open Data Week's prestigious Data Through Design exhibition, our team at Blockparty was honored to be selected as one of ten exhibiting artist teams, showcasing innovative applications of open data for public good. We created a compelling art installation that visually mapped housing inequality across New York City, integrating data analytics results from our AI NLP platform.
Architectural design projects from 2015 to 2020, from generative design, community engagement, to urban tech.
No projects match the selected filters.