Nils Helset is a 15th generation Norwegian farmer who embraced cutting edge technologies to gain the best yields from his land and build a successful agricultural business – DigiFarm.
We live in the information age and farming is no exception! In children’s books a farm usually has one cow, one sheep, a horse, and a few chickens and ducks. In the 21st century farming is much more industrial, technological, and challenging. Climate change, population growth, soil degradation and other large-scale global issues are now everyday problems for farmers.
Scientific innovation is key to assisting those tasked with producing our food and protecting our ecosystems to make optimal decisions about their processes. Accurate data about soil moisture, precipitation levels, crop yields, and optimal use of seed protection chemicals, are just some of the things that modern farmers need to know to become more sustainable and produce with precision.
DigiFarm brought a successful agricultural software product to market by combining first-hand knowledge of farm practices with the compute power of EuroHPC JU supercomputer LUMI. By delivering workable solutions for real world agricultural problems, DigiFarm has gone from strength-to-strength since their inception in 2018 and has picked up multiple HPC industry awards along the way.
Field of information
Farming with precision helps to reduce costs and boost crop yields, thereby driving efficiency overall. While farmers walk their land every day and know it very well, the foundation of precision agriculture lies in having highly accurate field boundaries.
Until recently, existing solutions for large-scale field boundary data were outdated, inaccurate, and time-consuming to update. Landowners had to manually draw field boundaries into digital farming platforms. And this was a sizable task! There are approximately 34.6 million field boundaries in Europe and over 32 million in the US. These were drawn by hand in 2008 and have not been updated much since then owing to the extremely time consuming and resource intensive nature of this work. A 5-hectare field takes 1 person approximately 1 minute to manually draw; 66.6 million boundaries would take 125 years. As seeded acres change every year, these datasets grow inaccurate quickly and provide poor data for farmers trying to conduct analysis or deliver services.
DigiFarm allows now farmers to see their land from a new perspective by enhancing satellite imagery with Artificial Intelligence (AI) and supercomputing technologies. This bird's eye view of their holdings, mapped with digital accuracy, provides farmers, and the agriculture and food industries, with information they can use to support data driven decision-making.
Down on the digital farm
Where farmers previously had to rely on out-of-date maps, lived experience and self-declarations, they can now access custom made technologies to support their important work. DigiFarm co-founders Nils Helset and Konstantin Varik developed a state-of-the-art deep neural network model. The model is super-resolving high satellite imagery data from the European Space Agency, Sentinel-2 at 1 meter per pixel resolution and utilising 4.7 million hectares of training data from 57 countries. This deep neural network model is optimised to provide highly accurate measurements of field boundaries, planted areas and crop health and yield, both in-season and historically. With this model, DigiFarm achieves detection accuracies above 94%, which is 12-15% better than existing methods. The solution reduces agricultural monitoring costs by over 25%, saving Lithuania, for instance, €1.5 million annually.
In the early development stage DigiFarm received essential support from the Norwegian Competence Centre for HPC who are part of the National Competence Centre network that is co-funded by EuroHPC JU. The National Competence Centre network (EuroCC 2) provides national contact points to foster the uptake of HPC by industry, and other strategic communities. The LUMI User Support Team (LUST), who provide comprehensive and targeted services to new HPC users, including support with workflows, tools and skills, were also instrumental in getting DigiFarm up and running when they started out. In June 2024, DigiFarm were awarded 35,000 computational hours by EuroHPC JU on LUMI, to further their research and development.
Access to LUMI accelerated DigiFarm’s ability to train its deep neural network models. It reduced the time spent on research and development including the testing of new models, adjusting hyperparameters (crucial to controlling machine learning processes) and loss functions (differences between the perceived and actual outcomes of the model). This early intensive work was critical to DigiFarm’s ability to commercialise solutions and models for their clients.
The awarded hours on LUMI also allowed DigiFarm to increase the accuracy of their models. This had a significant impact on DigiFarm’s competitive advantage and opened new geographical markets in agricultural regions with multiple growing seasons such as the Asian Pacific and Africa. With a team of more than 40 employees, DigiFarm is now serving more than 50 corporate clients including ENI, Bayer Crop Science and Syngenta.
In 2025, DigiFarm is adopting the latest technologies like transformers, deep learning architectures used in machine learning that can assist with image classification, object detection and image generation for computer vision. They continue to develop their core research and development infrastructure through ongoing partnerships with EuroHPC JU and LUMI.
Cream of the Crop
DigiFarm’s work with LUMI resulted in a HPCwire Editor’s Choice Award for Best HPC Collaboration 2024, as well as the 2024 HPC Innovation Excellence Award from Hyperion Research (HPC User Forum). These prestigious awards recognise the value and excellence of this European product.
DigiFarm perfectly demonstrates what can happen when European innovation and infrastructure come together to push the boundaries of what’s possible. They have BIG plans to continue creating world leading products in the field of agriculture, remote sensing and Artificial Intelligence.
Working with EuroHPC supercomputing infrastructure allows startups like DigiFarm to accelerate their development stage and absorbs some of the costs of expensive, technical infrastructure that is often beyond the means of emerging enterprises. Nils recalls the early challenges of acquiring and maintaining large-scale computing systems, which proved both difficult to house and expensive to run. Today, EuroHPC’s infrastructure, including its 13 new AI Factories, is designed to assist European SMEs, scientists and startups by providing customised supports and compute power for free.
If you would like to hear Nils Helset talk about this project, you can access the EuroHPC 2025 Summit Live Stream on Showcasing Scientific and Societal Breakthroughs on EuroHPC Supercomputers:
DigiFarm was awarded computational time through the AI & Data Intensive Applications Mode:
