NSF & Atlas AI: improving the science for agricultural efficiencies

Anaïs Tadlaoui

For more than 40 years, America’s Seed Fund powered by the National Science Foundation (NSF) has helped startups and small businesses like Atlas AI transform their ideas into marketable products and services. Since 2019, our work has been bolstered by the support of this incredible program and this month we are thrilled to kick off our work as a NSF Small Business Innovation Research (SBIR) Phase II recipient.

Atlas AI co-founder, David Lobell and our CTO, George Azzari, have been pioneering the application of satellite-survey methodologies to improve access to timely, reliable data on agricultural productivity in data-sparse environments, (such as) in emerging markets, for more than a decade.

Our global agriculture systems fuel and sustain all of us — it touches every aspect of our lives from the food we eat, the clothes we wear, the cotton sheets we sleep in, the cars we drive. The agrofood system creates millions of jobs and is critical in achieving sustainable development goals. However, information gaps in African agriculture prevent industries from growing. Some existing forecasting methods for cereal crops use precipitation and other remotely sensed indicators, but they cannot produce accurate, reliable predictions from year to year. Others use field data to model yields, but these approaches rely heavily on the collection of ground-truth data and are therefore not scalable.

With the support of the SBIR program, we aim to advance state of the art technology by adapting satellite-based deep learning methods and biophysical crop modeling for the complex farming systems that are characteristic of sub-Saharan Africa.

More specifically, as NSF SBIR Phase II recipients, we are working to commercialize a new technology for producing high resolution agricultural yield forecasts for the heterogeneous farming systems found in most emerging economies. Our objectives are to (1) develop in-season crop classification methods to identify where cereals are being grown; (2) implement pixel-based yield forecast models for a broad range of geographies.

Our work over the next two years will advance basic knowledge in the field of satellite-based crop area and yield prediction, particularly at the field-level (for which no methods are currently established). The resulting findings will be published in a series of blog posts and peer-reviewed scientific publications. We will also be using the technology to develop software products that can be used to improve operating efficiency for the banks, insurers, agribusinesses, and other firms that service smallholder farmers across Africa.

We’re truly grateful for the continuous support of the NSF in helping us further our research in aim of providing insights into sub-Saharan agriculture, a key driver to intra-African trade.

To learn more, you can reach out to us here.

Sign up for our newsletter to get the latest news on investing and growing your business

Subscribe
world map illustration