Russian Agritech Firms Develop AI “Digital Breeder”

Russian photoseparator manufacturer SiSort and researchers at Voronezh State Agrarian University have developed an experimental AI service for agronomists.

The neural network was trained on the works of Sergey Goncharov, a renowned plant breeder, Doctor of Agricultural Sciences, and professor at the university’s Department of Breeding, Seed Production, and Biotechnology. The system preserves his communication style by incorporating his personal notes and correspondence into the training data.

Supporting Agronomists

The project aims to make academic expertise accessible to practicing farmers. Users can consult a virtual version of the scientist about crop varieties, cultivation specifics, and common challenges. The assistant helps guide decision-making under uncertainty, taking into account weather conditions, market dynamics, and other critical factors.

The AI service currently does not process images. Developers plan to integrate it with another SiSort tool, the Kalibr bot, which measures seed parameters from photographs. In the future, agronomists will be able to photograph a problem, have Kalibr analyze the image, and receive recommendations from the digital “professor” based on its knowledge base.

SiSort is also developing a digital catalog of weed species in Russia in collaboration with the university. The service, accessible through the Kalibr bot, allows users to identify weeds by photographing them on a white sheet of paper with a coin for scale. The system can recognize the species and, in the future, will be able to forecast their impact on crop yields and recommend herbicides. The current database includes about 160 samples of major weed species and is expected to expand. The company has also automated the generation of analytical reports on agricultural markets using big data.