This Nairobi startup is using AI to add data science to African agriculture

Farmers need data to monitor and predict everything that goes into affecting their crops from drought and flood conditions to soil health and temperature fluctuations.

But for many—particularly those in developing nations—such facts and figures aren’t readily available. The problem is especially worrisome in Africa, where some 60% of the population is engaged in small-scale farming. “Africa is still the most data-scarce continent,” says Kate Kallot, CEO and founder of Amini, a Nairobi-based environmental data startup founded in 2022 and backed by investors like Salesforce Ventures and the Female Founders Fund. Much of the data is siloed in paper files in government offices or only available through prohibitively expensive satellite providers, she says. And global data sources and models are often less accurate in Africa than elsewhere on the planet.To fill in the gap, Amini is collecting environmental data in Africa through technologies like satellite imagery, drones, and IoT sensors, as well as from existing studies. Her company then analyzes the raw data using artificial intelligence, and provides actionable information and recommendations to farmers, crop insurers, farm lenders, and governments to optimize agriculture on the continent.

Amini announced a deal late last year with Aon and the African Development Bank to use its data to promote affordable crop insurance across Africa, but the company also operates infrastructure to send more immediate notifications to individual farmers. Amini can send automated texts to farmers notifying them of situations like impending floods or spreading pest infestations, and it’s increasingly using AI to be able to automatically respond to texted queries about weather and other conditions.

“That’s the beauty of technology,” says Kallot. “It’s very complex at the back end, but at the front end, the only thing they’re getting is a text saying, ‘be careful, there will be two weeks of extreme rainfall the next couple of days.’”

The data and guidance can help farmers grow their crops more sustainably—and make it easier for them to obtain crop insurance, loans, and other financial services to make their businesses more sustainable. (It can also be beneficial to crop buyers looking to work with farmers using sustainable agriculture practices.)

Amini has been working with HP and Nvidia (where Kallot previously worked as head of global developer relations), which have provided workstations and GPU processing power that have enabled Amini to build efficient AI systems in Kenya. In one case, using Nvidia’s GPUs rather than ordinary CPUs sped up computation by a factor of 23, Amini machine learning engineer Clinton Oduor said in a recent talk.

Building locally, rather than offloading computation to overseas cloud computing systems, saved Amini a lot of money and allowed the company to employ engineers and data scientists on a continent that Kallot says talented developers often feel compelled to leave for better opportunities.

“A lot of them, their grandparents, or their parents, are farmers or pastoralists, so they know exactly the challenges that their parents and their grandparents have been facing,” says Kallot.

Working with Amini goes toward HP’s efforts to have a sustainable impact, says Jim Nottingham, senior vice president & division president of HP’s Advanced Compute & Solutions business. It also offers proof of what HP customers can accomplish with the company’s technology, he says, with workstation-based AI not just valuable in places lacking local big data centers or having limited connectivity.

“This is making a difference for the planet, for the people and for the communities,” he says. “And we see this as something that can scale.”

AI developers often begin experimenting with workloads on local computers even if they’ll later move them to the cloud, he says, and some organizations will only use workstations or private cloud setups for privacy and security. Over time, HP has created workstations outfitted to start doing AI work essentially out of the box, equipped with development toolkits for working with Nvidia GPUs, Python environments favored by AI programmers, and a “Stack Manager” tool for quickly installing the right versions of software for a particular task.

Working with AI can seem daunting after hearing reports that big models like OpenAI’s GPT can take thousands of chips and months of work to train, suggested Bob Pette, vice president of Nvidia’s professional visualization product group, in a recent panel discussion that also included Kallot. But smaller AI models for specific purposes can be developed on affordable workstations.

It’s likely Amini’s technology will also come to have an impact even beyond Africa. The company has had talks with teams in Barbados, Brazil, and Southeast Asia about how the technology can help address local problems.

“Supporting local teams with the right computing infrastructure to then start understanding the challenges in their communities is what’s next,” says Kallot. “So we are kind of enabling other Aminis to also seek to also solve the problems of their communities and their countries.”