Over 1.2 billion people around the world lack access to electricity, most of them in rural areas of Sub Saharan Africa and Asia. An accurate map of infrastructure is critical to expand access to these people. Yet, the electricity network is often under-mapped and the schematics that do exist are not publicly available, or lack geospatial accuracy.

In this talk, we'll describe our work with the World Bank to map the grid network in OpenStreetMap in an efficient and repeatable manner. We developed an approach using machine learning to identify areas likely to contain high-voltage towers and used this information to speed up human mappers more than 10-fold. We will highlight this strategy of augmenting human effort with machine learning and provide perspective on how this workflow might be applied to other challenges facing the mapping community.