Food isn't just about taste, it's a fascinating intersection of biology, chemistry and data collection. In this series, we are exploring the research shaping what we eat and why, from breakthroughs in nutrition and sustainability to using to technology to improve crop yield.
When Mariano Alvarez arrived at Duke University as a postdoctoral researcher, he never imagined he would run a seed company. But the origins of his startup, Avalo, can be traced directly back to Duke's biology department, its greenhouses, and a few faculty members who guided him toward an unexpected future.
Alvarez spent nearly four years as a postdoc in Duke biology, where he studied how plants respond to their surroundings, how they develop new traits, and how those traits are passed on to future generations.
His work involved both evolutionary theory and computational genetics, which led him to the office of Duke computer science professor Cynthia Rudin .
He remembers walking in "like a lost puppy," but Rudin welcomed him.
"She was nice and helped me think through some machine learning approaches that might help with my genomics work," says Alvarez.
He implemented those ideas, they worked well, and that's what eventually spun out into the company. Avalo is using machine learning models to accelerate the development of new, resilient crop varieties. They are currently working with cotton farmers in Texas and sugar cane farmers in Australia. The company also has been testing broccoli and tomatoes in its greenhouses.
The biggest challenge in making a seed is knowing ahead of time how it will perform in real conditions, including in different climates, soils and regions. Machine learning helps to forecast performance before planting.
Traditionally, farmers select desirable crop traits by identifying plants that already show those qualities, such as drought resistance, and using them to pollinate others. The seeds are then grown and tested in the field. However, because each plant must complete its full life cycle before results are known, the process can take many years.
AI, on the other hand, can predict the performance of a seed without having to grow it, speeding up the process by as much as 70 percent.
"Because of Cynthia's work, we can understand why a seed will do well, for example, which genetic variants help it tolerate heat. AI also helps us plan business decisions by simulating product performance and cost. None of that would have been possible in any reasonable timeframe without AI," Alvarez says.
Rudin recalls her early conversations with Alvarez. "I'm excited about the impact he is making," she says.
Rudin's work has long focused on the tools Alvarez needed. "Sometimes, we want to know which variables are important for a specific predictive model; other times, we want to know which variables are important in determining the outcome," Rudin explains. "This is important for genetics, because we want to know which genes cause something to happen."
A Startup Born From Late‑Night Conversations
The idea for Avalo didn't emerge in a lab meeting. Instead, it developed over beers with his best friend, co-founder and CEO Brendan Collins, who also has a strong biology background.
They took the idea to what is now Duke's Office for Translation & Commercialization , which helped them spin out the technology in early 2020.
"For the first year, we were basically just two guys with phones and a computer," he says.
Support from Duke didn't stop there. His postdoc advisor and biology professor, Kathleen Donohue , was supportive and encouraging, even when his work drifted outside her expertise. Duke's mentorship program also paired them with Diane Ignar, "the first person to explain how we should go about the process of starting a business," says Alvarez.
Alvarez now realizes how unusual Duke's support was. "Duke's spin-out terms were pretty good and pretty clear… it really matters for startups, and Duke was the best-case scenario for us."
Traditional breeding still relies on crossing plants and waiting to see which offspring perform well. Avalo replaces that guesswork. "We do a little 23andMe on them," Alvarez explains, referring to the genetic testing company. "We sequence each of their genomes and generate reports that tell us which ones are going to do well in farmers' fields ahead of time."
To accomplish that, Avalo has become a fully functional, small seed company that operates out of the Alexandria AgTech campus in Research Triangle Park. It has about 10,000 square feet of greenhouse space spread across RTP, Duke and NC State. It handles everything from breeding to computation to field trials.
Why Cotton - and Why Now
About 80 percent of Avalo's work focuses on cotton. The choice wasn't arbitrary. The company evaluates new crops through three lenses: technical feasibility, economic potential and sustainability impact. Cotton hit all three.
Texas is the largest producer of cotton, contributing approximately 40 percent of U.S. cotton production in recent years. But with water becoming increasingly scarce, farmers are facing a crisis. "A lot of the growers we talk to in Texas don't have water … they're dependent on the rainfall that they have and those are really variable and really unpredictable conditions," says Alvarez.
That reality is echoed by the farmers with whom Avalo is working, as well as those who want to work with them.
Andrew Hurt is among those considering working with Avalo. A fourth-generation farmer from Plainview, Texas, Hurt recently toured the hot houses behind Duke's biology building, where Avalo is growing cotton.
"Every generation ahead of me, technology has had a big part in the efficiency and success of that operation."
Fourth-Generation Farmer Watches the Water Disappear
But the challenges he faces today are different from those his father and grandfather confronted. Earlier generations focused on water conservation. "We farm in a very hot, very volatile area … and declining water tables," Hurt says.
That's why Avalo's approach caught his attention.
"Having the ability through AI to know and quantify better these varieties and how they can work in our area is extremely valuable," says Hurt.
Roger and Rachel Wade, who farm cotton, peanuts,cattle, hay and some wheat, have been through a season with Avalo.
They joined the company's grower program last year, "and everything went really well," Roger Wade said. The economics stood out immediately. "We made a pretty decent crop last year, and we got a little bit more money for it than average market value."
In a region where rainfall is disappearing, that matters. "With these seeds, they don't need as much water… that's kind of the theory, and I think that may be correct," Wade adds.
In March, the company announced that the first cotton crop produced more than 30 percent more than projected.
As for the role of AI in agriculture? Roger Wade echoed the questions many growers share: "It's very interesting. I'm sure there are lots of good things that it can do. So, we'll just have to see what comes of it in the future."
A Growing Footprint - From RTP to Australia
The company recently raised $11 million in Series A funding and has partnered with Coca‑Cola Europacific Partners. That partnership supports Avalo's second major crop: sugarcane in Australia. The funding is going toward scaling up its commercial programs, increasing its footprint in Texas, and expanding operations in Australia.
Even after leaving Duke, Alvarez's ties remain strong. He continues to receive help from his previous advisors, and several employees came to him from Duke.
And the greenhouses - where Alvarez once conducted his own research - still house Avalo's plants. "It's really cool to have our plants… sitting next to these experiments in the same greenhouses," he said. "It's just kind of satisfying to see."
Alvarez sees Avalo broadening into more crops and more acres. "Ultimately … the impact of the company is best realized when we can get our technology to as many people as possible."
