Flox Intelligence Uses AI to Keep Wildlife Away from Trains, Airports, and Roads

Preface

When a train hit a wild bison on the way to Mumbai, the collision damaged its engine and left passengers stranded for hours. Such incidents are not confined to one region: wildlife encounters can disrupt transport systems and put both animals and people at risk around the world. Aviation and rail figures show that the impact can be substantial, from costly aircraft strikes to widespread train delays. After years working in transportation and autonomous vehicles, Sara Nozkova concluded that simply making traditional deterrents louder was not the answer. She helped develop a different approach: artificial intelligence that recognizes nearby wildlife and selects responses suited to the animal and its behavior. This article explains how Flox Intelligence’s technology works, where it is being used, how the company plans to grow, and what challenges remain as machines move beyond controlled environments and into nature.

Lazy bag

Flox Intelligence uses on-site AI devices to identify nearby wildlife and emit sounds intended to deter animals without relying solely on fixed barriers or generic alarms. Its system aims to adapt responses so animals do not become accustomed to repeated signals. The company has deployed about 150 devices across several regions and types of sites, and also sells software for monitoring and reporting. Flox’s longer-term ambition is to extend its technology to vehicles and other machines operating outdoors. Its growth plans include a $10 million Series A to expand deployment and develop applications beyond current wildlife use cases.

Main Body

Encounters between wildlife and transport infrastructure can have consequences far beyond a brief delay. A personal example helps illustrate the danger: when Sara Nozkova was eight, a train traveling to Mumbai came to a sudden stop after striking a wild bison. The collision damaged the engine badly enough that it had to be towed away for repairs. The passengers resumed their journey about eight hours later. The incident was memorable, but it also pointed to a broader issue: animals and infrastructure can come into conflict in many parts of the world, not only in rural areas.

The financial and operational effects can be significant. The FAA estimates that wildlife strikes cost U.S. civil aviation about $272 million a year on average between 1990 and 2025. In France, rail operator SNCF says wandering animals affect more than 8,000 trains across the country each year and rank among its five leading causes of delays. In Brittany, animals were hit by 120 trains in 2023, disrupting 660 services and causing more than 15,000 minutes of delays. These figures describe disruptions to transport, but the underlying problem also raises concerns about the safety and welfare of animals.

Traditional ways of keeping wildlife away include sirens, lights, fences, falcons, and culling. Nozkova’s view is that a more effective solution should respond to the animal and situation rather than simply intensify a fixed signal. “I was convinced the answer was not a louder deterrent,” she told TechCrunch. Her perspective was shaped by 13 years of work in transportation and autonomous vehicles. That experience showed her how systems designed for predictable environments can struggle when natural settings introduce changing conditions and living behavior.

Machines often perform well in places built around known layouts and objects, such as cities, factory lines, and warehouses. Nature is different: animals move unpredictably, surroundings change, and conditions are rarely fully controlled. Nozkova describes this gap as a limitation in how machines perceive the world. Systems that navigate engineered environments may become unreliable outdoors because the setting does not follow fixed geometry or a consistent set of expected objects. Flox’s central idea is to build a layer of perception that helps machines interpret wildlife activity rather than treating nature as an unknown obstacle.

To pursue that idea, Nozkova joined the AI lab at KTH Royal Institute of Technology in Stockholm. There, she began working on a system that could vary its deterrent response according to what a nearby animal was doing and what kind of animal it was. The work later became Flox Intelligence, which spun out of KTH in 2020. The company is among the finalists in this year’s Startup Battlefield 200 pitch competition at TechCrunch Disrupt, scheduled to take place at Moscone West in San Francisco on October 13-15.

Flox’s devices process data at the site rather than sending it to the cloud for analysis. This is known as edge processing. The approach resembles the perception systems used in autonomous vehicles, but Flox trains its models to recognize wildlife behavior. A motion sensor detects movement and wakes the device. A camera then captures the nearby scene, and on-device models identify the animal. Based on that identification, the unit emits a sound selected to be appropriate for the species and intended to scare it away.

The system is also designed to learn from experience. Rather than playing the same signal in the same way every time, it can vary its responses, with the aim of reducing the risk that animals become habituated to repeated deterrents. This adaptability is central to the company’s approach: a response that takes the species into account may be more useful than a generic alarm, while variation may help preserve its effectiveness over time. The goal is to discourage animals from entering hazardous areas, rather than to injure or remove them.

Flox currently sells battery-operated units that can be mounted on poles, trees, and other structures. Its software platform gives clients an overview of where devices are deployed and what is happening at those sites. The platform can also produce reports for regulatory compliance and support other applications, including biodiversity monitoring. In this way, the system is intended not only to deter wildlife but also to provide information about activity in the surrounding environment.

The company’s longer-term objective is to make its software portable enough to run on machines that operate outdoors. Potential platforms include trains, cars, and drones. That ambition reflects Nozkova’s broader argument that physical AI is moving beyond factories and into complex outdoor settings. Machines such as trains, trucks, drones, tractors, and robots may need to recognize and respond to natural surroundings as part of their operation. Flox presents wildlife intelligence as a possible additional layer in the perception systems used by such machines.

Nozkova says about 150 Flox devices are currently live across North America, Europe, and Australia. Sites include airfields and airport perimeters, road and rail corridors, national parks, agricultural land, urban areas, mining sites, and critical infrastructure such as electrical substations. Along transport routes, the company particularly identifies fence ends and crossings along wildlife corridors as relevant locations. These are places where animals may encounter openings or crossing points near infrastructure, and where awareness of wildlife activity may help operators manage risks.

Flox’s commercial model combines equipment sales with recurring software revenue. Customers purchase edge units up front and, after the first year, pay an annual subscription for software and AI. The company also has original equipment manufacturer (OEM) partnerships with businesses that pay an annual fee to deploy Flox software on vehicles such as trains. Train manufacturer Alstom is one of those partners. This arrangement could allow the technology to be integrated into equipment rather than relying exclusively on separate devices installed beside a route.

Nozkova said Flox expects to end 2026 with revenue of about $1.5 million. The company operates in a market where other systems also use technology to manage wildlife encounters. Robin Radar’s Scarecrow uses species-specific distress calls to disperse birds and records activity. iCHASE detects birds with computer vision and targets them with a moving laser, adjusting its response as needed. These examples illustrate different approaches to the same broad challenge: identifying wildlife and influencing its movement near places where encounters can cause harm or disruption.

Nozkova says Flox’s potential advantage lies in the breadth of its system and the data it has collected in the field. She describes this information as “ground truth on what actually makes a moose or a grizzly bear move safely.” Field data matters because a system must work with real animals and real environments, not only controlled demonstrations. However, the company’s long-term proposition also depends on whether its devices and software can perform reliably across different species, locations, and operating conditions.

Flox has raised about $5 million across its pre-seed and seed rounds. The seed round was led by Unconventional Ventures and Tilia Impact Ventures. Co-investors across the rounds include GENIUS NY, Norrsken, Satgana, and angel investors, according to Nozkova. The company is now opening a $10 million Series A. It expects most of the new funding to support growth from about 150 devices to several thousand deployed units, as well as extending the underlying system beyond its current wildlife applications.

Flox’s approach does not remove the need for careful evaluation. Deterrents must be appropriate to the species and setting, and operators need to know whether a system is reducing dangerous encounters without causing other problems. Deployment across airports, rail corridors, agricultural areas, and critical infrastructure also presents varied practical requirements. Still, the company’s work addresses an important limitation in outdoor automation: machines built for controlled environments may need better ways to perceive living, changing surroundings. If Flox can demonstrate that its adaptive system works reliably in the field, its technology could serve both as a wildlife deterrent and as a foundation for machines that interact more safely with nature.

Key Insights Table