In Latin America and the Carribean, countries face stark disparities in children's health between neighboring countries; and chronic undernutrition greatly impedes the livelihoods of children and their future outcomes.
In order to secure the livelihoods of every child, it is imperative to generate up-to-date data to measure the impact of public policies related to food security and household conditions. Tracking the growth of cities and informal settlements complements this data to give a big picture.
Using our analytics platform, companies, governments, and NGOs can access up-to-date, ready to use data to make better decisions that impact their populations.
We hope to continue releasing more tools as open source.
Over the course of the year, we have iterated several times to test new features and business models. The first release was a web map where users could order layers of imagery and receive the data in a matter of hours. This strategy proved to be an unsustainable business model as users in some cases only required imagery without analytics.
We have also tried other models, like API access, or preprocessed indexes like floods or soil use; however, this strategy doesn’t work to scale. Finally, we shifted our approach to a machine learning as a service model (aka MLaaS). The key point of this model is to match with companies that already have imagery and want to accelerate the development of tailor-made machine learning models, and also match with people that want analytics for a specific area. For this service, we charge users a monthly fee that provides hours of computing, storage, and serving of their models. This is our current business model and we will keep testing it.
Our commitment to Open Source has remained steadfast, and our algorithm of informal settlements mapping has significantly grown and now contains use cases for countries in the region like Argentina, Paraguay, Guatemala, Honduras, and Uruguay. For example, our team worked with the government of Uruguay on deploying algorithms to analyze their own imagery.
While creating algorithms, we faced some issues associated with the lack of technical tools to process imagery at a large scale, so we released three internal tools as open source: dask-rasterio for parallel imagery processing and two QGIS plugins for batch zonal statistics extraction.
The next steps are based on the roadmap of Dymaxion Analytics — we aim to form partnerships with drone/satellite imagery and agriculture companies. For now, our partnership with the NGO Techo was renewed for the year, currently focused on mapping in other districts of Argentina. We plan to expand to other countries since our tool has been validated by users and generated revenue in Argentina.
Last but not least, we would like to thank the UNICEF Venture Fund team for their support in the business, technical, and, most importantly, personal support to make the Dymaxion Labs team better entrepreneurs. We learned a lot during this year of the program and encourage other startups to apply.