All cases

AI-powered fungal infection detection system: Protecting ecosystems through early intervention

Client
Nonprofit organization specializing in global climate control research
Industry
Biotech/Environmental Science
Services
Software development, AI
Tech stack
FixMatch, TensorFlow, OpenCV, Python, OpenFIBSEM

Challenge

Our client, a nonprofit organization dedicated to global climate research, focuses on early detection of fungal infections in plants to prevent ecosystem-wide outbreaks. Their existing process relied on highly trained bioinformatics specialists to manually collect and analyze samples—a slow, resource-intensive approach that hindered rapid response efforts. To scale their mission, the nonprofit sought an AI-powered solution that could automate fungi spore detection and classification and recommend a containment strategy while remaining accessible to field workers without advanced technical expertise. The system also needed to integrate seamlessly with their lab equipment—an electronic microscope. They were looking for a vendor with solid expertise in AI development who could also work with hardware.

Our team’s responsibilities:
The challenge lay in balancing scientific precision with simplicity—creating a robust AI system that could democratize fungal monitoring without compromising accuracy or speed. To achieve this, our team performed the following tasks:
Built a cross-platform application that interfaced with the microscope the client used for sample analysis
Developed computer vision algorithms to automate microscope control (zooming, focusing, and scanning samples)
Trained semi-supervised machine learning models to identify different fungal types
Designed an intuitive user interface for non-experts, including visualization tools like fungal pollution maps
Implemented real-time analytics to track infection progression and recommend containment actions

Results

We built a cross-platform application that works on Windows and Linux. It also interfaces with the electronic microscope, which is connected through USB. Computer vision algorithms are directing the microscope to move and zoom in and out. Our solution democratizes sample analysis by allowing someone without specialized education or training to walk through an area and manually collect biosamples, such as fallen leaves, into specialized tubes before placing them into the device that runs our AI-powered app. Next, this AI system works as follows:
Computer vision algorithms direct the microscope to optimize sample scanning without human intervention
If an infection is detected, TensorFlow models classify infections and predict spread rates using temporal data (e.g., changes between weekly samples)
The system builds fungal maps, visualizing microbe colonies on plants and surrounding areas
It offers pollution tracking functionality as it generates heatmaps of infected areas across regions and timestamps outbreaks
The AI-driven recommendation engine suggests containment measures (e.g., targeted pesticides) based on infection type and spread patterns
Our solution can scale monitoring efforts by allowing researchers to analyze samples twice a week, or more frequently if needed.
Fungi monitoring system
Fungal monitoring software

Impact

The solution we implemented delivered transformative results, including:
Offered 10x faster sample processing time
Enabled non-experts to conduct accurate fungal detection with less than one hour of training, dramatically expanding monitoring capabilities
Identified and classified 50+ fungal strains using computer vision and machine learning models
Generated real-time containment recommendations, allowing field teams to implement targeted treatments faster
Integrated seamlessly with existing lab equipment, maximizing the client’s technology investments

Latest projects