Leveraging Machine learning, soil and weather data for sustainable agriculture

65% of farming is conducted through traditional methods passed through oral tradition. While these methods have sustained us for generations, the unprecedented climate shifts, changes in the food markets, wars and market expansion require us to adjust our techniques to fit the contemporary conditions. 

For instance, Zimbabwe this year had a late-season rainfall. More so  El Niño resulted 25% reduction in annual rainfall, thus exposing 2.8 million people to food insecurity risk and malnutrition rates rating to to over 5.7%. Challenges like that are not native to Zimbabwe alone, they are experienced everywhere periodically and they spread quickly burdening neighbouring countries.

Agriculture as a sector also suffers a great deal from malpractices that are hazardous to the environment. T

Without producing a surplus for contingency, environmentally conservative methods and data-driven approaches the future of Agriculture for many countries remains bleak.

While some of these challenges in Agriculture are inevitable a lot could be done to alleviate the problem.

 We seek to improve produce and create surplus starting by using classifier models to predict crop diseases. We'd then use Large language models to produce detailed prescriptions and courses of action for farmers. In addition, we have a data connection with ISRIC soil grids and present area-specific data in easy-to-interpret graphs. That would be important in implementing environmentally sustainable practices for both soil management and disease control. We also intend to decentralise weather data. We have intertwined our systems with the open weather API  and chained the data with Large Language models such that a farmer gets forecasts and detailed suggestions on rightful practices. 

Our secondary solutions include communities around the platform, inputs-produce market/supplier connections and robust project management systems that extend to finance handling and more.

Our solution will be especially useful for small to medium-scale farmers. In most cases, they do not have funds to send samples to expensive laboratories for soil analysis, moreover, they can't pay for expensive weather data services and finance software. 

Our platform will increase farm produce directly bolstering their economic output. Additionally, our users will adhere to practices that conserve the environment and ensure sustainable agriculture for their communities. Ifta will also facilitate networking with suppliers and buyers as a secondary layer that ensures farmers get the best deals 

Our team is growing and has yet to fully realise its full form. Currently, I work as the tech lead. I have experience in machine learning, programming and web development. My partner Kudakwashe Foya has experience with social engineering and has been impactful in various ventures, which include mentoring social impact entities in partnership with prestigious organizations like MIT Grand Hackathon, Merge and Peace first. Currently, she serves as  a Director Of Partnerships at Fatima Al Fihri Open University and a Regional ambassador with Peace First. Her work in social impact has gained her strategic national, regional and global recognition a social impact leader, as exemplified by her Emy Africa award nomination in 2023 (https://shorturl.at/rvHU7), The Varkey Foundation and Chegg as a Global Student Prize Finalist(https://t.ly/PNBiJ), The Diana Award(https://rb.gy/xgml0h), UNICEF Innocenti(https://shorturl.at/dsyQY), Aspire Institute(https://shorturl.at/bmqFI), and Crisis UK(https://shorturl.at/fqzJZ), and recently as an awardee of the African Woman In Development Award  by Donors for Africa(https://shorturl.at/qANST).

Her experience and network coupled with my technical skills enable us to meet stakeholders at their pain points. I for one, was born of small-scale farmers in Mt Darwin, Zimbabwe and for nearly 11 years I have seen their struggles. Consequently, I am very knowledgeable in terms of the knowledge gap that exists within farmers of different scales. I've also experienced working with community Agriculturalists deployed by the governments, and I'm cognisant of the lack of scalability of such a system as their help often requires them to be present at various sites. 

My colleague Kuda holds regular meetings in numerous communities, especially as an ambassador of Peace First Zimbabwe. This puts her in well position to disseminate information to numerous audiences seamlessly. That is key in our organisation as a chunk of our task will involve regular fieldwork, workshops and educational programs for farmers to use technology independently. 

We intend to grow our team, however we are careful to do this on a need basis. And as we proceed we'll get more technical team members. 

• We have made all necessary API connections
• We have working models to predict diseases
• We have prompt-engineered GPT-4 for comprehensive prescription
• We have a working web interface
• Project management systems framework is functional
• We have an existing marketplace framework
• We have created a system that allows different User types (Farmers, Suppliers etc)

We are hoping to expand our team and cover computing costs. And that would require funding, more so we'd be interested in doing field research and surveys to know we can fully meet our customers at their point of needs. 

In addition to that, training AI models is hardware intensive, so we would like to work towards getting computers or cloud computing resources that ensure that. 

Last but not least, we believe we would benefit a great deal from being part of the MIT solver community. That would come in the form of knowledge/mentorship, networking and meeting like-minded people from the community. 

The majority of prevailing Agritech solutions make use of expensive hardware to mine data for specific farm data. For instance, the use of remote sensors, geo-surveying equipment and satellites for weather data sourcing. In addition, farmers may need to hire professionals to interpret data the data, maintain hardware and find the implications. 

While this may facilitate highly accurate data, it is not affordable and scalable. In small to medium economies, that luxury is only afforded to a minuscule number of elite farmers. The aftermath is the majority of the land held by smaller and less equipped farmers remains underutilized. 

Our solution provides data and comprehensive explanations with little to no extra hardware costs from the farmer's side. For instance, we have robust and reliable API connections to ISRIC (International Soil Reference and Information Centre), which implies via our platforms farmers can access soil composition insights.  Therefore we can provide farmers with details about Nitrogen content, organic carbon content, pH levels and more, for areas at least 10km^2. 

Furthermore, we have extensively created prompts that allow GPT-4 and other LLMs to create recommendations that foster sustainable and effective land use. We have the same for weather data via open weather API. 

In addition to this, we have custom classifier models, that enable disease prediction. The predictions are further processed using large language to provide real-time prescriptions and course-of-action recommendations.

Overall, our solution decentralises farming data to the masses without the need for expensive professionals, hardware and laboratory. 

At the start of the process, farmers get access to the IFTA platform. From there they can make well-informed decisions, and agricultural practices from land preparation, seeding, pest & disease control and harvesting. Farmers are also afforded project management tools and a marketplace as a secondary tool for sourcing inputs and potentially selling products. 

The results would include better land conservation, bolstering of crop yields and sustainable finance decisions. Our overall objective is to economically empower farmers as they ensure everyone has a plate of food on the table worldwide. 

We have three impact goals in a broad sense. These include

1. Food security
2. Sustainable Agriculture and environmental conservation
3. Economic empowerment for farmers

1. Food security 

At the forefront, we strive to contribute towards having every child have enough nutritious meals daily. That extends to households. And the only way to get to this is by scaling national agricultural produce. Farmers should be equipped to sustain growing populations in various parts of the world, especially in Africa where poverty is ravaging. 

2. Sustainable Agriculture and environmental conservation

Poor agricultural practices are responsible for soil erosion, river siltation,  deforestation and high levels of soil & water toxicity. What drives farmers to counter-productive methods is often a lack of information. Farmers often have limited knowledge about their soil, prevailing weather conditions and disease control. Consequently, they default to environmentally unfriendly methods. We would like to solve that by providing information on demand.

3. Economic empowerment for farmers

In lesser developing economies farming is a profession that does not attract talent. Often relegated as a means for those who felt they had no other lucrative option. That in turn hurts the industry. The reason why that is, is most farmers are not economically empowered. Often they are viewed as inferior citizens even though they bring food to everyone's table. Therefore by increasing their yield and bolstering their business, we live to see the day farmers are not in abject poverty and become respected member of society. Further more, in most countries, a large share of the population comprises of farmers. Therefore moving farmers from poverty would significantly increase the overall well-being and standard of living for the host country.

Our core technologies include Machine Learning, IoT and Web technology.

We use Machine learning mostly in disease identification through image classifier models. Additionally, we use Large language models to provide comprehensive explanations and prescriptions.

The Internet of Things is responsible for communications through APIs to our data sources for soil and weather. 

Web technology provides User Interfaces and facilitates interaction between the user and the system. At the backend, the technology also ensures requests to numerous APIs. 

Overall much of the technology we use has been used in other systems, thus its tested and proven.

Derby Tendai Matoma (Full time), Kudakwashe Whitney Foya (Full time)

2 years

Our team is fairly small. I am an African young man aged 25, my teammate is an African woman aged 25 too. As we progress we'd love to have a lot more young people of all genders from different parts of the world. 

Our primary product is well-processed data, to enable data-driven decisions among farmers. We deliver this through the Web, and later on, we'll develop mobile versions of the Web app. This data ranges from soil composition data to weather and custom disease control information based on images. Our data is set in a way that makes stewardship-minded decisions easy. In turn, farmers can get bountiful harvests, and successful businesses while conserving the environment. 

In terms of revenue, we are to follow a subscription-based model. With monthly and annual subscriptions available for all users.

While we understand we can't get thousands of users overnight our penultimate goal is to support the business through user subscriptions. As a secondary means, we'll also sell APIs to our models. 

Currently, our development stage has been heavily subsidised by Microsoft for startups and open AI, who offered us 25,000 USD of cloud computing resources. These resources can take us through a year or more, enough time to get the product out.

During the early stages soon after launch, we may heavily bootstrap the project while seeking external funding. We expect this to last for 8-12 months. There we should be in a position to self-sustain selling our products to farmers as well as our disease prediction system's API to other developers solving similar problems.