Meet Sophie Zeng, the 17-year-old California student who built an app to identify invasive plants after learning they can become overlooked wildfire fuel

Wildfires are a familiar threat in Southern California, but when Sophie Zeng began looking more closely at what helps some fires spread, she became interested in something much smaller than flames or drought: plants growing quietly across the landscape.

The California student learned that some invasive plants can alter ecosystems and, in certain landscapes, add readily combustible vegetation that contributes to wildfire cycles. She also saw another problem. Many people who encountered these plants might not know what they were looking at, much less how to remove them.

So Zeng turned to artificial intelligence.


At 17, she was recognized as a 2025 runner-up for the Gloria Barron Prize for Young Heroes after co-founding Project ARISE, an environmental organization centered on biodiversity restoration. A major part of the project is an AI-powered mobile app designed to recognize invasive plants and provide information on how they can be removed.

California's wildfires led her to an overlooked problem

Project ARISE began in 2023 after wildfires affected Zeng's home region of San Diego County.

According to the Barron Prize, Zeng began researching invasive vegetation after learning that uncontrolled invasive plant growth can contribute to wildfire fuel loads. She concluded that one barrier to removing invasive plants was surprisingly basic: identification. People cannot easily remove a problem plant if they cannot distinguish it from the native vegetation growing beside it.

The wildfire connection is backed by a substantial body of ecological research, although the relationship varies by species and ecosystem.


The U.S. Geological Survey says invasive grasses that burn more readily than some native vegetation have increased wildfire frequency in Southern California shrublands. Fire can then clear native shrubs, allowing invasive plants to occupy the newly opened ground. That vegetation can create conditions favorable to subsequent fires, producing what scientists often describe as an invasive grass-fire cycle.

In other words, fire can encourage certain invasive plants, and certain invasive plants can help create conditions for more fire.

She trained an AI model using thousands of plant photographs

Zeng decided to attack the identification problem with a tool increasingly familiar to teenagers: a smartphone. But making a phone recognize invasive plants required considerably more than taking a photograph.

The Barron Prize says Zeng trained her own deep-learning algorithm to classify some of California's most invasive plant species. She assembled thousands of photographs, including images she took herself and others collected from digital databases, to create the data needed to train the model.

Deep learning is a branch of machine learning in which artificial neural networks learn patterns from large amounts of data. For image classification, a model can learn visual characteristics: shapes, textures, edges and other features, that help distinguish one category from another.

In Zeng's case, those categories were plants. According to the Barron Prize, her model eventually reached 90% accuracy. She then led a student team working to integrate the algorithm into a mobile application. The goal is not simply to tell a user the name of a plant. Project ARISE says its app is intended to classify invasive species and help users understand how to deal with them.

Why invasive plants can make wildfire problems worse

Not every invasive plant increases wildfire danger, and invasive vegetation is far from the only factor driving destructive fires. Weather, drought, ignition sources, topography, land management and native vegetation all matter.

But certain invasive grasses can change how fuel is distributed across a landscape.

USGS researchers explain that species such as cheatgrass can grow quickly, dry early and leave behind a carpet of dead material that burns readily. A 2026 USGS review and meta-analysis similarly concluded that many non-native invasive grasses increase wildfire activity and can recover quickly after fire.

Southern California provides particularly relevant examples.

The University of California's San Diego County invasive-species guidance identifies giant reed as an invasive plant that can contribute to higher fire frequency and intensity. UC material also notes that non-native annual grasses can dry before the summer and fall fire season, leaving combustible plant tissue behind.

That helps explain why identifying vegetation can be more than a botanical exercise.

Project ARISE moved from an app to pulling plants from the ground

Zeng's idea did not remain confined to software. Project ARISE began organizing invasive-plant removal events, habitat restoration projects and environmental workshops, working with park rangers and conservation organizations. The group also launched initiatives focused on native pollinators, including distributing mason bee houses.


The Barron Prize reported in 2025 that Project ARISE had reached more than 60,000 people and removed an estimated 18,000 pounds of invasive plants. The organization's website now puts its estimated total at about 19,500 pounds, based on reported volunteer service hours, with more than 1,300 hours of member service contributed.

Project ARISE has also expanded beyond San Diego. Its website describes chapters across multiple locations and says its broader mission is to use technology, education and hands-on restoration to protect biodiversity.

At 17, Zeng turned computer science into conservation

Zeng's interests do not fit neatly into one field. Project ARISE identifies her as its founder and adviser and notes her long-standing interest in programming and artificial intelligence. She is now an incoming freshman at Stanford University.

Her project also illustrates an unusual use of AI. Instead of asking an algorithm to generate something new, she trained one to help people recognize something already growing around them.

The app alone cannot prevent wildfires, nor can removing invasive plants eliminate California's fire risk. The science is much more complicated than that.

But Project ARISE addresses a smaller, practical question: What if more people could recognize potentially damaging invasive vegetation before it spreads?

For Zeng, answering that question started with thousands of photographs and an algorithm. It eventually put volunteers on the ground, pulling thousands of pounds of unwanted plants from California landscapes.

FAQs

What does Sophie Zeng's invasive-plant app do?

The Project ARISE app is being designed to use artificial intelligence to recognize and classify invasive plants. The project aims to help users identify invasive species and learn how they can be removed. Zeng trained the underlying deep-learning model using thousands of plant images.

How accurate is Sophie Zeng's AI model?

According to the Gloria Barron Prize, Zeng's deep-learning model achieved 90% accuracy when classifying the invasive plants it was trained to recognize. That figure describes the project's reported model performance; it should not be interpreted as meaning the app can identify every invasive plant everywhere with 90% accuracy.