Saathi - An AI innovation for Nepali farmers
Despite being a country driven by its agricultural sector, Nepal has been suffering from negative growth in food crop production. This major issue affects both the population and the economy, resulting from a culmination of drastic climate change, degradation of soil and a lack of information among the farmers. The majority of farmers still follow experience-based, conservative farming methods – as a consequence, they are left unable to keep up with the changing agricultural factors.
But what if the farmers had meaningful access to scientific agricultural methods to make the most of the land? Having timely scientific information on which crops to farm and how, rather than making do with the knowledge currently available to them, would be a game-changer.
Impacters launched Project Saathi (“a friend” in Nepali) to conceptualize a product that could help Nepali farmers adopt data-driven farming with ease. Saathi is an innovation that automatically collects, analyses and communicates data to the farmers. It is designed to provide simplified advice throughout the planting process, from crop selection to maintenance. It empowers farmers by reducing the need to seek and wait for consultation. Instead, they can receive support in real time, from choosing which crops to plant, to monitoring their health.
About the project
Impacters wants to encourage young women in STEM by creating opportunities for them to gain tangible experience in computer science, rather than solely practising advocacy for their inclusion in STEM spaces and providing theoretical training. This allows them to better relate to the field and see the results of their work. Making meaningful and scientific contributions to agro-domains encourages the team members to keep exploring related fields and topics, and provides them with the opportunity to become role models for future innovations.
Project Saathi brings together an important, contemporary and local problem, namely Nepali farmers’ lack of access to scientifically informed practices, and the team of young scientists ready to mobilize their education and gain hands-on experience. On top of the obvious benefits to the smallholder farmers, Saathi serves as an incubator for multiple groups of our female innovators to mobilize their Impacters training in Machine Learning and Robotics as they tackle the issue at hand through self-directed research and experimentation. Within the scope of Project Saathi, our team of young scientists and engineers was tasked with designing both hardware and software to create an inexpensive, non-intrusive and easy-to-use system that empowers farmers through scientific data-assisted agriculture.
The hardware side of the project focuses on creating a prototype that is able to perform edge computations, has little to zero recurring expenses, is easy to set up and is rugged to withstand difficult environments. In order to address these requirements and still perform within the limitations such as power storage, limited resources etc., a Saathi device consists of two different units.
Sensor unit: This standalone component, powered by solar energy, is meant to go into the farm and continuously collect data from the soil, then communicate with the second component.
Processing unit: The second component is a portable component with a screen. This is the processing unit that processes the data collected from the sensor unit, analyzes it and presents it in a way that the farmers can understand.
The sensor unit stays on the farm, while the processing unit is carried around by the farmers. The processing unit receives the sensor unit data without relying on an internet connection and the data is automatically synced every time the two components are in close proximity.
Apart from this inter-unit data exchange, the Saathi devices themselves can share encrypted data with each other using a peer-to-peer network. This ingenious decentralized protocol allows Saathi devices to share data without a centralized cloud infrastructure.
The most important component of our software is Artificial Intelligence (AI)/Machine Learning (ML). To give the farmers essential recommendations and alerts about their crops, we are using AI and ML to process soil and plant imagery as well as the data points collected by the sensors from the soil. The ML computations are done on the edge device (processing unit) itself, eliminating the need to connect to the internet.
The on-site lack of internet access presents an interesting challenge given that ML computations are very resource-intensive. In general, a small device like a smartphone cannot do large computations by itself; the processing is usually done on much more powerful infrastructure on the cloud and the processed data is sent back to the device. This approach of sending the data to the cloud and getting the processed information back is not feasible when working with farmers who cannot use sophisticated devices and some of whom are living in places without an internet connection.
This project surmounts this challenge by creating innovative and efficient edge solutions that allow computations to be executed without the need for the internet. The device is equipped with various in-device ML models that continuously gather and interpret information about the farms. As a result, the device can stay with the farmers themselves, enabling them to make continual use of AI without need for the cloud.
Tailoring the interface to the user
One very important part of the software is the user interface (UI) that makes sense of the information interpreted by the processing unit. Suitable UI design is crucial in catering to non-tech savvy farmers and to making things as easy as possible for them.
In terms of the workflow, our UI is very simple so that any farmer can use our product with minimal supervision. The software is very intuitive, ensuring that the users do not have to understand the complexities of the system to make full use of it.
In essence, the scientific data are collected automatically and the farmers only need to provide a couple of photos of the soil or the crops. Saathi acts like an assistant by collecting data, running it through our system and outputting very easy-to-understand information to the farmers.
Our target groups
Farmers remain our primary audience and our product is designed to solve a critical problem the farmers are facing. As it stands, most of the farmers currently follow the information passed down through generations without any scientific background. A very few aware farmers, who are open to adopting scientific assessments in their farming pratices are limited to annual soil testing through the use of mobile labs, vans or soil laboratories for soil testing. These infrastructures require very detailed technical knowledge that the farmers do not have. In practice, this means that the barrier between the end user and scientific assistance has not been solved. Furthermore, these infrastructures are too few and in too high demand during the appropriate windows before planting season, rendering the actual help the farmers get from these testing ineffective. Saathi solves this problem by providing a self-contained system that is simple to install and use and that continuously monitors the farms and provides suggestions in real time.
Saathi is not only helpful to the farmers; it also serves the researchers seeking solutions for collecting accurate scientific data from the farms. Besides its primary focus to provide simplified summaries to the farmers, Saathi is a powerful machine that collects and stores extremely useful data — to researchers/scientists, these data lead to important agricultural insights. By supplying regular information straight from the field, Saathi can truly kickstart next level scientific discoveries and innovations in the field of agriculture.
The future plan
In order to stimulate further research and development of this solution, this academic prototype is being open-sourced, rendering it available to a variety of public and private actors who see potential in the progress that our team has achieved over ten months. Saathi tangibly demonstrates the possibility of delivering timely, up-to-date and suitably formulated scientific advice directly to the farmers. This example stands to be expanded and iterated upon by the open source community even after the completion of the project, providing a new paradigm geared toward actively empowering the target demographic.
Beyond the specific product and its potential, this type of project is meant to drive interest in researching and developing new strategies to assist farming and promote scientific assessment of agricultural methods. Saathi exemplifies how investing in research and development can further the agriculture sector through technology.
Impacters has created a large network of female tech leaders who are regularly trained to build technological solutions to solve tough societal problems. This network includes, but is not limited to, current and aspiring computer engineers, electrical engineers, mechanical engineers, agriculture scientists, UI/UX (user experience) engineers and data scientists. Project Saathi unites these female tech leaders to innovate in the field of agriculture in Nepal in a space that includes meaningful support and supervision from established businesses and experienced specialists.