Research
Big Data &
Aid Targeting
In Myanmar, “Using Artificial Intelligence/machine learning to evaluate the distribution of community development aid across Myanmar” (2025) examines how community development projects are distributed spatially across local communities. Specifically, this research seeks to understand which communities are more likely to receive large amounts of development aid based on their poverty levels and other development characteristics.


Contextualized Poverty Targeting
AI/ML models are predominantly developed in countries with georeferenced national surveys. “Contextualized Poverty Targeting Through Multimodal Spatial Data and Machine Learning in Brazzaville, Congo” demonstrates that household-level targeting can be achieved in a data deficient city without a gold standard survey. Using the case of Brazzaville, this study integrates intuitive images, social media, points of interest, connectivity, and administrative data to predict multidimensional poverty at the household level. The simulation in this study shows that a spatially augmented targeting method not only reduces targeting errors but also decreases the poverty ratio, gap, and severity.
Developing
AI-Informed Vulnerability Index
Beyond model performance, a critical gap persists in the application of these predictions to policy design. Zambia multimodal project examines how AI/ML based poverty prediction outputs (generated at a 1km-by-1km grid cell) from multi and unimodal data can be applied to guide resource optimization. This research leverages actual budget allocation of food aid across the finest geographic boundaries in Zambia). The research translates the prediction into concrete budget reallocation based on poverty reduction objectives.


Detecting Human Interpretable Features to Support Poverty Mapping
“Detecting human-interpretable features from satellite imagery to support poverty mapping” focuses on detecting human-interpretable features of poverty, such as different types of roads (paved vs. unpaved), water bodies, and buildings, with pixel-level accuracy. This state-of-the-art semantic segmentation model utilizes sub-meter resolution images from PlanetScope and represents the first vectorized map of poverty features in Africa.
The Last Mile in Remote Sensing Poverty Prediction
For high-stakes aid allocation decisions, achieving near-perfect predictions across all communities is crucial. “Last Mile in Remote Sensing Poverty” study applies explainability techniques (such as Integrated Gradient and Guided Grad Cam) to evaluate standard vision models (Vision Transformers and CNN architectures such as Resnet and VGG16) for their poverty prediction tasks. We then enhance predictions with multimodal architectures and investigate cases where all models struggle.

Geotargeting in Urban Areas Informed by
Community-Driven Insights
This paper explores how to use geospatial data to identify urban households experiencing food insecurity, which is more challenging to predict than asset-based measures. Specifically, we introduce ways in which community input can be integrated into our poverty predictions in Lusaka, Zambia, with the objective of addressing the lack of local appropriation that often results from geographic targeting or scoring approaches.
Climate Change and Internally Displaced Persons
“Assessing the resilience of climate-induced internally displaced persons “ examines internally displaced persons (IDPs) affected by climate change in the Philippines. The project aims to analyze factors related to community resilience, such as social capital, disease exposure and awareness, and government support, among IDPs in Tacloban and Batangas.

Global Poverty
Creating accurate indices of extreme poverty is an important ethical, empirical, and policy-relevant consideration. Credible and reliable measures of poverty are powerful instruments in the drive toward greater justice and toward securing the rights of all individuals to a decent standard of living. The ways in which poverty is measured shape perceptions of need and have repercussions for targeting interventions.

Empowerment
of Women
Nari Gunjan’s work with the Musahars of Bihar has been an exemplary case of a women’s empowerment program that has targeted an extremely needy group. As the case study has revealed, the significant increases in literacy rate among girls at the NG centers and awareness of their well being is a testament to the important contributions NG has made to the Musahar community.

Enhancing Poverty Measurement
Traditional poverty measurement based on surveys does not provide granular data on income/consumption, wealth, or development at many specific locations of interest, thus limiting effective interventions targeted at communities. Dr. Jung's work aims to address this critical gap.