Mingjia Zhang (’29)

This summer I worked in the World Bank Group’s DIME (Development Impact Evaluation) AI Lab, where I contributed to ZeroHungerAI, a project utilizing machine learning to improve early warning systems for food crises in developing countries. My work focused on building and evaluating alternative models for forecasting food insecurity levels across more than 80 countries using data on climate, conflict, food prices, vegetation, and socioeconomic factors, as well as features derived from news reporting streams. I explored different model architectures and representation learning techniques, evaluated how well models generalized to countries with limited historical data, and worked on making predictions more useful and understandable for decision makers.

Working at the World Bank showed me how technical research and theoretical concepts from my classes are applied to problems with significant real-world stakes. I found working at the intersection of machine learning and international development particularly insightful; building a better model in this context requires especially careful consideration of data limitations, interpretability, and how predictions will actually be used. The experience gave me a much clearer view of what it takes to develop reliable machine learning models and strengthened my interest in developing systems that can help address large-scale challenges.