Olds College students present AI research at conference in Brazil
Published on Friday, July 10, 2026 at 10:45am
Olds College of Agriculture and Technology students present their artificial intelligence research in Brazil.
Zachary Komarniski, a third-year Bachelor of Digital Agriculture student, will speak about his work on machine learning for precision agriculture at the 17th International Conference on Precision Agriculture, held in conjunction with the 11th Precision and Digital Agriculture Conference, July 11-18 in Porto Alegre, Brazil.
In collaboration with Olds College lecturer and researcher Felipe Karp, Komarniski’s project focuses on revolutionizing the way the agricultural industry cleans and processes large geospatial datasets. This project was made possible by a Mobilize grant from Olds University, which funded research and development of the framework.
“The ultimate goal was to train the machine to perform advanced human-style data analysis,” Komarnisky said.
“What would take a human over 10 hours to manually review all observations to remove anomalies can be accomplished in just seconds with our machine learning models, greatly increasing the ability to move into the analysis phase of research projects with clean data.”
Modern precision agriculture relies heavily on vast amounts of data generated by equipment and sensors. However, this data is often subject to inaccuracies due to operational inconsistencies, statistical anomalies, and other outliers. As Komarnisky points out, any inaccuracies negatively impact the final analysis, making data cleaning and filtering an important step for both producers and researchers.
Currently, that data filtering process typically requires manual analysis and highly complex filtering and parameter tuning. For data collected from a single field, this means humans can spend minutes or even hours reviewing over 100,000 data points. Komarnisky’s research aims to change that.
Using data from the Hyperlayer Data Concepts project, Komarnisky and Karp developed a machine learning framework that automates the detection and removal of invalid or anomalous data while preserving relevant data. By splitting large datasets into chunks and feeding them to a model, you train computers to perform human data analysis in seconds, ensuring the cleanest data possible without accidentally removing valid data points.
Komarniski’s presentation in Brazil will focus specifically on evaluating the accuracy of this machine learning framework on complex geospatial data.
“There is a talent problem in the agriculture industry that is rarely talked about. There are very few people who speak both the language of agriculture and the language of code, and Zachary is one of them, and he is only a third-year undergraduate,” Karp said.
“The machine learning framework he developed requires a deep understanding of how precision agricultural equipment behaves in the field, what anomalies look like in context, and how to design models that account for that complexity. That combination of knowledge is something most professionals only develop during their graduate training or even working in industry.”
