Ideas & Insights / All Perspectives / Ideas & Insights

Charting the Future of Atlas AI

Abe Tarapani - Chief Executive Officer, Atlas AI

A few months ago, I jumped into the opportunity of a lifetime when Atlas AI founders David Lobell, Marshall Burke and Stefano Ermon invited me to join the company as its new CEO. It wasn’t a natural moment to make a career transition — my work at Incandescent was deeply rewarding, and we were facing the height of the Covid-19 shutdown — and yet this wasn’t a hard decision. The United Nations estimates that somewhere between $5–7 trillion per year will be required globally through 2030 to achieve the Sustainable Development Goals. Developing the analytical tools to ensure this massive investment drives down poverty, increases food security, and results in stronger economies will require immense technical innovation, academic rigor, and AI capabilities adapted for data sparse environments. I joined Atlas AI because there is no other team I’d bet on to execute this critical mission than the group I’m now able to call my colleagues. I’ve written this post to share more about the vision we’re putting in place to guide our work for the next decade and beyond.

I met David, Marshall and Stefano in early 2018 when they were first thinking about launching what would become Atlas AI. Having spent years on the ground in Southeast Asia and Sub-Saharan Africa developing solar power projects, and time later in my career promoting the application of innovative data technologies for development challenges, I understood the power of their research. Better and more frequent data on economic well-being, agricultural productivity and infrastructure access across the emerging markets offers a wide range of potential applications — from positioning infrastructure, to forecasting consumer demand, to better program targeting and economic impact measurement. I admired the Stanford Professors’ desire to make this data accessible, well beyond the confines of academic journals, to solve meaningful societal challenges. It was a huge validation of their vision when The Rockefeller Foundation, an organization with a shared mission — achieving transformative solutions to the world’s great inequities — stepped forward to be a founding investor in Atlas AI.

It has become evident to me that the potential for Atlas AI is much greater than the sum of the individual data sets that have been developed to-date. From our earliest days as a company we’ve been focused on pioneering new machine learning techniques designed for high quality observation of ground conditions in regions with insufficient data to support traditional analytical approaches.

We currently apply these techniques to derive three types of data layers:

  • Productivity of key economic sectors such as the agriculture sector;
  • Infrastructure access and quality, focused currently on the power, transport and telecommunications sectors; and
  • Population density and economic well-being, featuring our estimates of asset wealth and consumption down to the village level.

We believe that sustainable economic development will be driven by progress at the nexus of the above measures, as illustrated by challenges such as:

  • Where do you place a new cell tower to ensure commercially viable levels of demand, while at the same time reaching underserved populations?
  • How do you monitor the downstream benefits of an upgraded rural road network on reduced post-harvest loss or improved public health outcomes?
  • How can tandem investments in electrification and irrigation drive growth in the agribusiness sector while catalyzing new micro-enterprise?

Across the emerging markets the relevant data often don’t exist to inform data-driven answers to these questions. Furthermore, the complexity of these economic systems makes it difficult to rigorously link specific actions (e.g., placement of a mini-grid near a specific village) with target outcomes (e.g., increased equity of energy access, greater financial wealth).

Addressing the above challenges at scale will require new machine learning methods to be invented; the ethical collection and fusion of large quantities of satellite, digital and ground truth data; and the development of software tools to abstract the complexity of this technology for non-technical decision makers. This is the long-term mission that we are embarking on at Atlas AI — to monitor the drivers of economic development across the emerging markets so that financial capital can advance societal well-being. We envision a world where there is no misalignment between sustainable development outcomes and investment returns, where it is commonplace to monitor the impact of investments to ensure equitable economic growth, and where capital investment flows frictionlessly within emerging markets due to deeper understanding of these economic systems and their relationship with long-term societal well-being.

In 2012, Jeff Weiner, at that time the CEO of LinkedIn, published a blog post introducing the concept of the ‘economic graph’:

I believe the opportunity ahead of us is to build Atlas AI’s own version of an economic graph, one that clearly outlines the interrelationships between people, industry, infrastructure and socioeconomic outcomes so that the world can more clearly work to realize economic opportunity and a more sustainable future. I’m humbled and grateful to be joining this world class team of machine learning scientists, software product specialists and emerging market domain experts. I’m inspired by our mission. And I hope you’ll follow us along this journey.


This article first appeared in Medium on October 7, 2020, and is reposted with permission.