Bringing predictability to a climate-vulnerable agri supply chain

India has witnessed a remarkable growth in agriculture technology, fueled by advancements in digital innovation and increasing adoption of smart farming practices. Integration of technologies like artificial intelligence, machine learning, and remote sensing has revolutionized farming techniques, enabling precision agriculture and predictive analytics.

IBT: Cropin has significantly impacted various stakeholders in the agricultural sector. Could you elaborate on how your suite of products, powered by AI, ML, and remote sensing, facilitates this transformation across different stages of the agricultural value chain?

Krishna Kumar: If you look at the idea of bringing data and intelligence powered by machine learning, AI, remote sensing, and knowledge graphs, it was to see how we can make every farm traceable, predictable, sustainable, and profitable. Those are the four pillars on which we build solutions for machine learning or AI.

We work with B2B intermediaries who then work with their customers, who are the growers, whether it’s a consumer packaged goods (CPG) company, food processor, government, or developer agencies in the sector, who then work with the growers to either source the raw material or produce alongside them. Or if it is the World Bank and government, they’re working on climate-smart agriculture, or how to do natural farming, or how to advise the growers to become better at doing the crop, right? So our approach is to ensure livelihood, better yield and quality.

Currently, we operate in 103 countries with our customers. Cropin runs many programs in India, Africa, Southeast Asia, and other parts of the world, where this intelligence is used by growers and companies to ensure farm productivity, quality, and yield. We manage more than 7 million growers on our platform, collectively cultivating 500 different crops and 10,000 different varieties. It’s a very extensive platform deployed globally, supporting stakeholders in making the right decisions and ensuring predictability for farmers in advance, making it sustainable and improving farmers’ livelihoods by ensuring income.

IBT: How have you structured your solution in accordance with the needs of the farmers?

Krishna Kumar: In order to achieve this impact, we have built an industry cloud for agriculture, which consists of three layers. The first is the application layer, where there’s an interface for everyone to interact with the information and data, organizing all the economic advisories. So there is an interface for organizations to comprehensively look at all the growers, what they are growing, what the risks are, and how they mitigate those risks.

Then there’s an interface designed to work in the field on mobile devices, where growers, agronomists, or farm managers interact with these datasets. For example, imagine a farm manager from PepsiCo or the World Bank running a program. They go to a particular farm, open the application, and see details such as the grower, the crop being grown, and the crop’s age. They also see risks from a disease perspective and receive notifications about potential issues, such as late blight on a potato farm in the next seven to ten days. If the crop has reached a storage or tuber bulking stage, they may also receive predictions about heavy rain forecasted five days in advance.

This intelligence allows them to advise on agronomy and plan the grower’s next steps. Growers receive these notifications in their local language via WhatsApp, IVR, or SMS, depending on their phone or receiver type. This application thus reaches every stakeholder in the process across the supply chain, providing enough intelligence and data for informed decision-making.

Now, in order to accomplish that, we have a data hub where we gather all the data surrounding the farm, whether there is a sensor or not. So, if there’s no sensor, we utilize satellite data from the past 40 years, current data, and forecasted data for the next six months. Various satellite sources like Planet, Landsat, ESA, or whatever is available around that farm are utilized. This data is collected and structured for each farm on a time basis so that models can run on top of it and provide intelligence for users.

Then, there are the models themselves. That’s a model layer, the third layer, where we build AI models either to detect crops, predict yield, forecast diseases in advance, identify climate risks approaching the farm, or issue alerts for deforestation or sustainability-related practices such as tillage or cover crops happening on the farm. These models are run on every farm daily or at a set frequency, providing intelligence to everyone.

IBT: With partnerships spanning globally and digitizing millions of acres of farmland, what challenges did you encounter in scaling Cropin’s solutions, and how did you address them?

Krishna Kumar: As a company, we have built the product in India for the globe and it was built in India with the complexity we deal with – smallholding farmers and different crops in different regions. And then we had to take this solution globally because our customers were looking for the solution to be deployed in multiple countries due to their operations there.

Language was one of the barriers. So if you are in Indonesia, it’s Bahasa; if you’re in Maharashtra, it’s Marathi. If you are in Spain, then it’s Spanish, or Brazil is Portuguese. So how do you make the system? Bringing the localization from a language and unit point of view. Somewhere you call it acres, somewhere you call it “junta,” somewhere you call it hectare, so different; or somewhere you measure it in square meters. So all those changes you have to make. And then you have to build AI models that can scale. So if the model is running on potato in India, can you run this in Brazil or in the US with the same consistency and accuracy? And that has been quite a work we have had to do based on the knowledge graph and trillions of datasets that we have trained over a period of time and made these models very, very accurate for our customers and growers.

Scaling those models has been a lot of work, and we spent around seven years doing it the right way. So every crop, every variety is grown differently, and they interact differently with the environment and the practices, and those nuances have to go into the models as well. So there is a beautiful synchronization of AI and ML geospatial expertise, crop science, and then climate science. So you have to merge all these different technologies and expertise to build these models which will scale globally. Otherwise, it becomes a proof of concept and then it never scales. We have a Cropin AI lab that keeps working on sharpening these models, and have done a lot of investment in that direction.

The other barrier is from the grower’s point of view. If you go to Africa, technology adoption is not so high, and then those farmers are not so technology-savvy. So if you have noticed, we have recently launched Akshara. While we are giving everything on WhatsApp and SMS and local language, we wanted to make it simpler, break the language barrier, and make it more interactive in a local language, where you can ask questions and provide answers, very contextual in your language to solve the problem.

For example, a grower may ask, ‘I’m growing potatoes, and this is the location of my farm. Tell me what the risks are from the disease point of view in the next seven days,’ and you should be able to answer that, ‘Okay, there is an 80% chance of late blight, and if you see those signs, you can use this method to take control of it, right? So, or there is heavy rain predicted or there’s a frost predicted. So can you make it more conversational with transformer models, language models?’ And recently launched Akshara, is a micro model, and we compressed this model to four bits so that it’s not so resource-hungry and we can actually train it on very contextualized data. So we trained that model on nine crops, five countries for the Global South. And we released it to see how relevant it is.

And then we saw that it has a 40% better efficiency than the Chad J54 turbo. We accomplished this in a highly cost-effective way. Our Cropin AI lab team dedicated extensive effort to train these models. They faced a significant challenge as the data is primarily biased towards global patterns, and the structure of data in this region differs greatly. Therefore, they had to meticulously fine-tune the models to ensure they were contextually relevant.

IBT: Cropin’s approach is described as crop and geo-location agnostic. Could you shed light on how this adaptability has contributed to your widespread adoption and success in diverse agricultural landscapes worldwide?

Krishna Kumar: I think we adopted this approach from the inception, aiming to create an impact on the last mile, starting with the smallholder. Now, if you go with a crop-by-crop approach, it’s very time-consuming. You build something for potatoes, then you’ll need to develop something for rice, and it’s a lengthy process. So we decided to make this application crop-agnostic and location-agnostic.

We aimed to create a general capability on the platform that is independent of crop and location. This means you can take this application to Africa and deploy it for 30 different crops, and it will still be manageable. The models will be able to run on those crops very quickly because they understand the crop, location, and climatic conditions. That’s how we designed the models as well. So we focused on making Cropin crop-agnostic and location-agnostic because our vision is to make every farm traceable on this planet. We aim to utilize all this intelligence to ensure farms are predictable, sustainable, and profitable. To achieve that goal, we can’t limit ourselves to specific crops. That means deploying a solution for only a subset. And that’s the choice we made, and I think it has worked beautifully.

IBT: The impact of Cropin’s solutions extends to improving the livelihoods of millions of farmers. Can you share some compelling examples or success stories that highlight the tangible benefits experienced by farmers and other stakeholders through your platform?

Krishna Kumar: Sure, I can think of an example of the work we did with PepsiCo, Lays. They wanted to support smallholder farmers growing potatoes for them, spread across the country with different varieties. The problem statement was that due to climate change, there was high variability in quality and production. They wanted to ensure the quality and yield of the raw material while benefiting the grower by securing their income and livelihood. We deployed a solution on these growers in multiple countries.

The solution aimed to predict diseases in advance, anticipate weather anomalies like frost or high temperatures, and identify water stress on the farm. We also aimed to predict the stage of the farm to manage procurement and storage accordingly. These predictions helped in production planning and decision-making for the companies. The project was successful, resulting in a 25% increase in yield, an 80% reduction in crop diseases, and a 92% climate adaptability rate. Income for the growers improved by almost 100%, around $15-20 per hectare.

Similar projects were conducted with the World Bank, focusing on climate-smart agriculture to support smallholders in Bihar and MP, as well as in Southeast Asia with ADPC and GEF. The World Bank published reports showcasing the technology’s impact on last-mile growers, highlighting income increases of 25-30% and a reduction in farm losses by 18%.

IBT: As the agri-tech landscape continues to evolve rapidly, what innovations or developments can we expect from Cropin in the near future, and how do you envision the company’s role in shaping the future of agriculture globally?

Krishna Kumar: At Cropin, we are always thinking about the next five to 10 years. There are a lot of new pipelines in development. We are ingesting data from all sources—public sources, private sources, sensors, drones, satellites—and bringing that data onto the platform in a structured manner. We are planning to compute at the scale of our planet. For example, if you take a crop like corn or potatoes, can we compute for every country where the crop is being grown and build intelligence for that grid or plot?

Additionally, can we forecast the yield potential for the next season, considering future climate scenarios? We have our long-term climate view, and we are currently working on this with some customers. The problem statement is that the supply chain is becoming vulnerable due to climate change, and accurate prediction of production levels is challenging. Can we provide this information at the grid level?

For instance, if you are planning to grow potatoes in Maharashtra or Idaho, can we identify which grid is highly suitable for potato cultivation next season? This requires a future outlook on how the supply chain will be affected from a potato perspective or where to grow potatoes in a new geography.

This information could also be valuable for growers. If a grower is planning to grow potatoes next season on their farm, they can ask Akshara, our micro LLM for agriculture, about the risks involved. Akshara can provide insights such as the possibility of frost attacks or heavy rainfall during maturity, which could interfere with the crops. It may even suggest growing a different crop. This requires massive computing at the country scale and is part of our global sustainability program.

We plan to compute at the country scale for the top five or top 15 crops and build a sustainability view of the supply chain considering future climate change scenarios. This will help stakeholders make informed decisions, whether it’s policy-making or expanding into new geographies. We are investing heavily in micro LLM, which we recently launched, and building high-speed computing capabilities to compute at the country scale for any crop and geography. These initiatives are part of our larger plan and we’re actively demonstrating them to our clients.

---

Krishna Kumar, CEO and Founder of Cropin, has dedicated the past decade to addressing complex issues within the global agriculture value chain by developing the world’s first Intelligent Agriculture Cloud at Cropin.

He launched Cropin with a vision to revolutionize global agriculture by integrating deep technology with agronomy. Under his leadership, Cropin has empowered millions of farmers and hundreds of agri-businesses, contributing to the resolution of critical global challenges such as food security, climate change, farmer livelihoods, financial inclusion, and biodiversity conservation. His entrepreneurial journey has been both exciting and rewarding.