FFplus Interview with Sumod Nandanwar, founder and CTO at PerPlant
In this interview, we speak with Sumod Nandanwar, founder and CTO at PerPlant, an agricultural startup whose mission is to adopt sustainable practices while maintaining profitability, based in Denmark. Sumod discussed transformative trends currently shaping the future of precision farming, introduced how PerPlant is integrating AI technologies, edge computing, data analytics and AI models training in their workflow and explained how are they developing a user-friendly, accurate and timely weed control system, using the innovative AI and HPC resources within the FFplus Business Experiment.
Sumod Nandanwar, founder and CTO at PerPlant.
Copyrights: PerPlant
To start, could you briefly introduce PerPlant?
Sumod:
We are a tech startup based in Copenhagen, Denmark, on a mission to democratise access to technology for sustainable farming. Our goal is to empower farmers to adopt sustainable practices without compromising profitability.
We do this by providing farmers with AI-based assistance systems and augmenting their existing machinery — such as tractors, implements, and harvesters — with smart camera systems. These cameras analyse crop health in real time and provide actionable insights, enabling farmers to spray more precisely and efficiently.
From your perspective, what are the most transformative trends currently shaping the future of precision farming?
Sumod:
Precision farming is a broad field, but I would summarise the most transformative technological trends into four key areas.
First, sensor technologies. This is one of PerPlant’s core areas. Advances in sensor sophistication and availability now allow for more real-time data collection — monitoring soil conditions, crop health, and pest presence. The growing range of sensors will help us gather increasingly granular agricultural data.
Second, data integration and connectivity. The data collected by sensors like ours needs to be seamlessly integrated with other farm systems to provide holistic, actionable insights that farmers can use.
Third, artificial intelligence. Once the data is collected and integrated, AI plays a vital role in analysing it to generate precise insights, automate decision-making, and enhance efficiency across the farming process.
And finally, robotics. Robots, leveraging AI-driven insights, can perform farming operations autonomously — from monitoring to precise application. Together, these four trends are redefining the future of precision farming.
PerPlant develops innovative products and smart solutions for spotting, spraying, and seeding. How are you integrating AI technologies, edge computing, and data analytics into your work?
Sumod:
Let me explain briefly how we use our camera systems. As I mentioned, we provide AI-based assistance that enhances existing farm machinery with real-time intelligence.
Our camera systems not only collect large volumes of data but also process it instantly on the edge, enabling machinery such as sprayers and spreaders to apply chemicals precisely — at the right place and in the right amount — saving costs and reducing environmental impact.
We approach this in three main stages:
Edge computing: Our AI models run directly on the edge, allowing real-time weed detection and spot spraying without cloud connectivity — it all happens while farmers carry out their usual spraying operations.
Cloud data processing: The cameras also send data automatically to the cloud, where we process vast datasets — from satellite imagery to sensor readings — to detect patterns, predict outcomes, and provide farmers with proactive insights for better planning.
AI model training: We have sensors deployed across Denmark, the Netherlands, Germany, Sweden, and the UK. These gather data under varied crop, weather, and lighting conditions. Using this diverse, high-resolution dataset, we train AI models capable of supporting farmers in different countries, benefiting all users collectively.
You mentioned the FFplus Business Experiment. Could you briefly describe your use case and the innovative solution you have been developing within the project?
Sumod:
Certainly. Within the FFplus Business Experiment, our use case focuses on developing a user-friendly, accurate, and timely weed control system using AI and HPC resources.
We aim to eliminate broadcast herbicide spraying, which is environmentally damaging and economically inefficient. Instead, we’re building an AI-powered system that uses image recognition and real-time processing to detect weeds and apply herbicides only where needed.
This approach dramatically reduces herbicide usage, lowers costs for farmers, and minimises environmental impact — all while improving operational efficiency.
As a newcomer to HPC, could you share your experience with these systems? Which infrastructure are you using, and who is supporting you?
Sumod:
As a small startup — though one with smart people! — we initially had no hands-on experience with high-performance computing (HPC). It was a steep learning curve to understand how to configure models, use the infrastructure, and manage training workloads.
However, once we got accustomed to it, the results were remarkable. We are using EuroHPC resources through Leonardo, specifically the Leonardo Booster, under the AI and Data-Intensive access programme. This provided us with access to powerful GPU nodes, essential for training large agricultural image datasets.
Our technical partner, D-Cube, has been instrumental in building the technical framework during the initial phase. Together, we adopted a two-step AI approach:
Extracting high-quality training data from over a million unlabelled images using unsupervised learning.
Training a large-scale Vision Transformer model, which guides smaller edge-deployable models through knowledge distillation.
HPC resources have allowed us to train complex AI models around 50% faster than before, automate much of the process, and capture key field features across diverse conditions — an essential step in scaling PerPlant’s solutions.
Looking ahead, what specific business benefits do you anticipate achieving once this experiment concludes?
Sumod:
We expect several major benefits once the experiment is complete.
First, we aim to reduce barriers to technological adoption in agriculture. Adoption has traditionally been slow in this sector, so making our precision spraying systems easier to integrate into existing farm operations is key. This will help us scale faster and broaden market reach while providing significant cost savings for farmers by reducing pesticide use and minimising crop damage.
Second, accuracy and reliability are critical for adoption. Farmers trust technology only when it performs consistently under real-world conditions. By leveraging HPC-trained AI models, we’re substantially improving detection accuracy and real-time performance — which will, in turn, increase farmer confidence and uptake.
Ultimately, this approach reduces environmental impact, helps farmers comply with EU regulations on herbicide reduction, and supports the EU’s sustainability goals. The EU’s subsidy programmes further reinforce adoption, creating a win–win situation for both farmers and the environment.
Through this project, we’re moving closer to our goal of making farming more sustainable. I’d like to thank FFplus for this invaluable opportunity.
Finally, what advice would you give to SMEs considering applying for the FFplus Open Call to explore HPC and AI for their business?
Sumod:
My advice is simple: just go for it.
We had no prior experience handling large-scale image data or using HPC resources, but the FFplus team guided us throughout the process. It helped us not only build the technical capacity to use HPC effectively but also understand its true value — enabling even small companies to train large foundational models and remain at the forefront of AI innovation.
This is a fantastic opportunity for small and medium-sized enterprises to leverage HPC for developing advanced models, enhance competitiveness, and contribute to technological and societal progress.
So, yes — just go for it. It’s not as difficult as it seems, and the benefits far exceed expectations.
Sumod, thank you very much for this insightful conversation and for sharing your experience. It has been a pleasure speaking with you, and we wish you all the best in your future endeavours.
Learn more about this sub-project