How one AI startup helps farmers grow better crops

Image of two people use the Ceres Imaging mobile app to review scans of farm land.

Two people use the Ceres Imaging mobile app to review scans of farm land. (Image credit: Ceres Imaging)

There are 2.1 million farms in the United States producing goods that generated nearly $400 billion in sales in 2012. The question these farmers face is the same one that has vexed them for millenia: How to improve the quality of their crops and maximize their profits? The answer rests in better understanding the ground beneath their feet. To ensure a successful harvest, farmers need to consider a host of factors like soil quality, temperature, and potential diseases. You might not think this is something artificial intelligence, let alone photography, could help solve, but it’s a challenge Ceres Imaging wants to tackle.

The five-year-old company mounts proprietary cameras and sensors on hired aircraft and takes aerial images of farmland. The sensors use wavelength-based spectroscopy to reveal the soil’s moisture level and overall condition, while also measuring crop health. The analyzed data detects stress points, and anything else that might not immediately be recognizable to the naked eye. (Ceres initially tried using drones for the flyovers, but its cameras were too heavy.)

Image of a camera belonging to Ceres Imaging attached to a plane.

A Ceres Imaging camera attached to the wing of a plane, scanning farm land and relaying the data back to farmers. (Image credit: Ceres Imaging)

“We’re more than putting a camera on a plane, but about data collection and how you deliver that to farmers,” Ceres chief executive and founder Ashwin Madgavkar tells me in an interview. “Our original thesis is around wavelengths and what types of data can be processed in order to create a repeatable process to generate returns to farmers.”

Farming with better data

Madgavkar conceived Ceres from his experiences as an electrical engineer, as a consultant helping energy companies explore cleantech, from time in South America working in agriculture, studying sustainability, and observing his Stanford professors use “novel techniques” like NASA sensors to gather data. “Working in South America gave me first-hand experience on the challenges farmers face. Existing imaging technology wouldn’t necessarily work,” he explains.

Farmers pay a subscription fee to use Ceres and can upload maps of the property they want to be surveyed. Ceres will monitor other areas if farmers are willing to pay for it, to gauge whether issues affecting their land comes from neighboring crops or for other competitive reasons. The company’s proprietary algorithm determines which farms to scan, flight plans, and schedules. It has an algorithmic model similar to that of Uber, matching rented planes and fields together—Ceres does not own any planes. Madgavkar’s team has trained the AI for four years using ground maps from university libraries and Ceres’ own data.

Image of a scan of farm land generated by Ceres Imaging's camera.

A scan of farm land generated by Ceres Imaging’s camera that farmers receive. (Image credit: Ceres Imaging)

“If we do a flight above a field and based on the models we’ve built, we can detect stress points, or a lack thereof. Those stress points are often points of emerging diseases that we can see before the naked eye can. With a vineyard or orchard, we can detect water content and a linear line, which often is a broken drip line,” he says. “When you’re managing a field, you have hundreds of thousands of plants. It’s easy to see problems with one plant, but hard to know for the rest. The result is that people miss the bigger issues.

Ceres primarily works with farms owned by professional enterprises, agriculture businesses, and retailers. Currently, Ceres works mostly with vineyards and orchards, specialty crops that have a higher margin than others, but that make up a smaller percentage of total farms in the market.

Madgavkar wouldn’t disclose specifics but said that his company works with “hundreds of individual farms, covering hundreds of thousands of different acres” and mostly in the U.S., with a smaller number in Australia.

AgTech getting competitive

If Ceres’ ambition is any indication, there’s an opportunity for solving agricultural problems using AI and machine learning. The company announced today that it has raised $25 million in new capital in a round led by Insight Venture Partners, bringing its total investment raised to $31 million. Madgavkar tells me that the new investment will help Ceres expand into new crops, namely corn, soy, and wheat, which make up 75 percent of global farmland.

Image of CB Insights graph highlighting five use cases of AI and robotics in agriculture.

Five use cases of AI and robotics in agriculture (Image credit: CB Insights)

Madgavkar is not alone in thinking technology can improve agriculture. More than $800 million was raised by AI and robot startups targeting this industry from 2012 to 2017, according to CB Insights. Problems being addressed include making crop planting more efficient, satellite imagery by drone, predictive analytics, and more.

Focused on imagery, Ceres isn’t in a class of its own, contrary to what Madgavkar tells me—he believes the competition is the “status quo” and “90 percent of farmers don’t use imagery.”

Ceres competes against the likes of satellite imagery startups like Descartes Lab, FarmShots, OmniEarth, and Orbital Insight; and in-field monitoring companies like Agri Eye, DJI Innovations, Gamaya, and others. Ceres’ differentiator might be the ability to combine imaging and AI into a single, low-cost solution.

“Farmers are quite high-tech when it comes to biological and chemical models they use, but they haven’t really embraced technology” yet, Madgavkar says.

It’s more than likely that large farms — those with over $1 million in sales — get this message, but they account for just 4 percent of all farms. Meanwhile, three-quarters of the rest of U.S. farms gross an average of just $50,000 a year. And when these small farmers face the age-old question of increasing crop quality and profitability, it’s likely they, too, will soon be farming with AI.