In Conversation with Lachana Hada: Saathi’s path, from concept to project to product

As Impacters prepares to celebrate the wrap of Project Saathi, we have been putting out materials online to provide insight into the product and the process by which the team worked to conceptualise and realise it. As part of this series, NCBL/WODES’s student intern Anitha Szabo spoke with Impacters Co-Founder Lachana Hada, who also served as a Project Head on the project, to shed more light on Saathi and the journey of the team that worked on it, as well as Impacters’ future plans.

Note: This interview has been edited for length and clarity.

To start a bit before the project’s inception, how were the idea and basic objectives for the project developed? I would like to hear more of the background, especially as it combines two different initiatives from WODES (Women Development Society). I was wondering if there was anything in particular that you would like to highlight.

At WODES, a few years ago, we did a project on organic farming, where we were teaching the farmers to do organic farming. What we realised during our time there was that as it stood, organic farming was not able to generate enough yield for commercial farming. It was enough, maybe, to sustain the farmers themselves but it was not enough to actually start producing commercially. And even though the farmers had been provided with the knowledge for the specific crop in the specific time, it was difficult for them to keep track of all the environmental factors that are continuously changing. We, of course, can provide them the knowledge to start with, but it is difficult for us to constantly update them in terms of the changing environmental factors. So, we recognized a need for a system that could constantly guide the farmer so they can farm efficiently.

On the other hand, we have Impacters, an initiative by WODES, that is always looking for projects that can solve real-life problems, projects that the girls can work on so that they can relate to and which they can be inspired from. At Impacters, we have been working for a long time on holding different workshops and projects to get girls into STEM. Instead of just focusing on advocacy or giving them a theoretical background, we try to find a project that they can relate to, something that is more tangible and where they can see and feel the changes that they are making. This makes it easier for them to relate the knowledge to real-world applications and understand what they are trying to do. In our experience, this gives them more motivation to continue learning even after the end of the workshop. This is one of those projects focusing on the girls’ involvement in STEM, but at the same time, we want to solve the problems faced by the Nepali farmers. We want to show them how they can contribute to society and when we are solving these real-life problems, we are tackling real-life societal problems. They can see the impact they are making and I think that is really, really, important. That is how we decided to bring this team together: we had the problems that the organic farmers were facing and we also had a project that was focusing on problems that affect people’s livelihoods, so the basic idea of the project was developed by bridging them together.

Is there anything that either inspired this project or served as a jumping-off point?

The product we are creating is the first of its kind in Nepal, as far as we are aware. There are some pre-existing services, as far as the soil analysis goes. There are soil vans that go around collecting soil samples. There are also a few labs that help the farmers analyse their soils, but this sample collection has to be done in a very specific way. It has to be done during specific months, there is a very specific technique for how you should collect the soil. Then it takes the vans and labs a couple of days to analyse the soil and gather all the data about the soil itself. Considering that everyone is doing this right before they plant, and that this testing is to be done right before planting, the result is a bottleneck in the amount of incoming data that the vans or labs can process. It might be months after the planting season when the farmers actually get the result. With this project, we are trying to create a product that addresses this problem the farmers have. Soil testing also is not something that anyone has really thought about making more easily accessible to the farmers. They keep thinking, “Okay, we need to increase the number of labs, we need to increase the number of vans,” for example, but nobody actually has considered making it directly available to the farmers. We are trying to solve that problem and given the current approach, I think that it is very important for people to see the innovation that we are bringing to the table.

As you are introducing a new approach to the problem, would it be correct to say that the prototype is an entire proof of concept to the credit of a direct approach?
Exactly. There are international projects that do something similar, but not exactly what we are doing because we are touching on several different things that focus on the specific needs of Nepalese farmers. The international products that do exist are also way too expensive for the Nepalese farmers to be able to afford.

What end product did the project generate, in terms of hardware components and the software housed by each of them? The UI features and user-facing functionality would be interesting to discuss, as there is a very specific target demographic.

The end product that we will deliver is called Saathi, which is a complete package that contains both the hardware system and the corresponding software.

First, we have the software. The most important component in our software is Artificial Intelligence (AI) / Machine Learning (ML). We are using AI/ML to process the soil and plant imagery as well as the data collected from the soil sensors to generate essential recommendations and alerts for the farmers in regards to their crops. However, the important thing when talking about AI and ML is that we forget that it is actually very resource-intensive. Generally, a small device like your phone cannot do all the processing itself, so the processing is usually done on powerful resources on the cloud. What actually happens is that the phones collect the data and send it to the server. The server then saves the data, processes it and sends the processed information back to the device. This approach of sending the data to the cloud and getting the processed information back is not feasible when dealing with the farmers who cannot use sophisticated devices. It is expensive, but more importantly, it might not be feasible in general because they might not have internet in those places. What is interesting about this product is that we are solving that particular problem by doing the computation on the edge device, so that the complete system is supplied with the product itself and the farmers can continue to make use of AI even without the cloud.

That was an overview of the software. Then we have the hardware, which is again divided into two different parts. We have a standalone device that is supposed to go into the farm, which we will call the sensor unit. It is a standalone system with sensors that continuously collects data from the soil and then communicates with a portable device. 

The second component, the portable device, has a screen. We will call this a processing unit since it processes the data collected from the sensor unit. It analyses, processes and then presents the data in a way that the farmers can understand.

The sensor unit stays on the farm and the portable processing unit is carried around by the farmers. The two units are synced every time they are in a close proximity without the need to connect to the internet.

To follow up on the rundown of the product, can you go into slightly more depth by outlining the intended normal workflow of the user? While trying to visualise the process, I was not entirely sure how many scientific implements the hardware would include and whether it would be measuring things like, for example, humidity in the soil.

We want to make the entire process as simple and as intuitive as possible. The idea is that any farmer who gets it should be able to use it with minimal supervision. It should be very intuitive to them and they do not have to do any of the measuring manually. They also do not need to understand the sensors or the system or the hardware. All they want is the information about their field and their crops, and they want to be able to understand the information.

The normal workflow would be as follows. The farmer takes the sensor unit and puts it in their farm. The sensor unit then starts collecting the data. When the farmers are then ready to plant crops, they take a picture of the soil with the processing unit that they have. They are then presented with a list of suitable crops for their particular farm based on the analysis from this picture they took of the soil and the data collected by the sensor unit and they can choose what to plant in the field. Once the chosen crop has been planted, they can continue to use this processing unit to constantly monitor the plants and the state of the field. For example, they can monitor the plant’s current and predicted level of water and temperature, they can also take pictures of their plants to see if they are getting enough nutrients and/or have any disease. 

In essence, we are using the soil data and the plant images to generate easy-to-understand summaries, so that the farmers do not really have to do much. We are doing the difficult task for them and they are just provided with a set of instructions that is easy for them to understand and follow.

How does – or how will –  liaison, training, and delivery to the farms work? Who would be responsible for this?

What is important to remember is that we developed this product in a short span of 10 months and that this is not yet a commercial product. In its current state, it is an academic prototype. Instead of creating a product that is sellable at the end, we want to use the product we have created as a prototype to kickstart research in the field of agriculture and to serve as the basis for further scientific research, especially at the intersection of machine learning and agriculture. And as such, what we are planning on doing is making this research open-source. We will make all the results and all the technologies that we created in this project publicly available to scientists and developers. This ensures that the open-source community can continuously contribute to the improvement of the product we have created. Since it is open-source, we will have a lot of developers, maybe even companies as well as individuals who will continue to build upon it and we will also keep researching with other interested partners.

In that case, would the scope necessarily include — unless it already did — a sample of farmers who would be testing it on ground? Or was there already a UAT (User Acceptance Testing) phase?

We do have a small network of farmers that we want to test with, and we also have a farmers’ association, who has shown an interest in our prototype. The problem is, because this is just an academic prototype, we cannot mass produce it. So if we were to say that we want to distribute it to something like a hundred different farmers, that is not possible at this point. But of course, it is something that we can look into in the future. If our partners are interested, then it is something that we could do further down the line.

This question is especially for the UI team; how were they able to troubleshoot and assure usability?

We had professional UI/UX designers consulting the team in regards to the UI- and UX-making, ensuring usability. They held a two-day workshop on developing user interfaces for non-tech savvy farmers and they have been in constant touch and providing consultation to the team.

What has the project revealed or proven about other and continued needs in smallholder farming?

Like I said, there is still a lot to do and this is just the start. We still need many new technologies to support the farmers; to increase their yield, to support sustainable development, but also to help improve common agricultural habits, like raising the quality of food crops, reducing the use of chemicals or changing the way farmers select crops, etcetera. To make the needed technologies a reality, we will have a lot of projects that build on machine learning and other technologies to create solutions that will help the farmers.

I would like to ask if you have additional comments about anything that has been indispensable to the success of the project. On the other hand, are there any needs or assets that only became apparent in the midst of the project?

Oh, the problem! The problem that became very apparent in the middle of the project, which we were not expecting, was how difficult it was to do hardware-related projects in Nepal. The hardware that we wanted for this project was not easily available in Nepal; it was difficult to find the components and we also had issues with imports. Ranging from the GPU we needed to run our machine learning algorithms to the processors and sensors we were using, it was not easy to get all the necessary hardware and machinery in Nepal.

The following question is about the composition of the team. What different groups were there for different parts of the project? Perhaps you can explain where the different group members came from, both the students and professionals, how this affects the way the project was carried out, and where some of them were planning to go or what they wanted to do afterwards.

Our project was divided into four different parts. We had a group that was focused on developing the machine learning models to generate recommendations and predictions. We had a group that focused on UI to make the usability accessible, making it simple and intuitive to cater to the non tech-savvy farmers. Then we had a hardware group that focused on developing the hardware prototype and finally,  the agriculture group that was focused on getting agricultural data and providing consultation for the agriculture related topics. They also supported our other groups to understand the problem from our customer’s, which would be the farmers’, point of view because they have more experience working on the ground level with the farmers. Those were the four different kinds of groups we had and each member of these groups was chosen corresponding to their prior education, experience and interests.

Some of them were more experienced in the sense that they had finished their bachelor’s degree, others were still doing their bachelor’s. The essential team members were the girls who had at least started their bachelor’s degree in electrical engineering or agricultural science or computer science. We also had mentors with more experience, they had already finished their bachelor’s and were doing their master’s or already working. They were guiding and providing support to the girls who were working on the different parts of the project. Additionally, we had advisors with more experience professionally as well as academically share their insights and expertise with the team members.

Where do the members want to go? There are a few who found a new passion in something they did not think of before they worked on this project. We tried to keep analysing the performance and the interest of the team and helping them find the area that they might like and fit better. By doing that, we were able to find areas for them that they might not have thought about, that they actually excelled in and that they want to look further into. There are others who want to continue to learn about machine learning and others still who want to continue with their academics and go on to do their master’s degree. Schools in Nepal tend to focus mostly on the theories and you do not really have practical experience in the topics covered, so what we presented them with was an opportunity to try things out that they might have studied. And working on them in real-life turned out to be very different from what they were expecting it to be, so now they have a clearer direction of where they want to go and how they want to achieve their goals in the long run.

I want to bring up the approaches taken by the organisation, the program and/or the team, in terms of project management, implementation, communication and internal training. Did any of these have to change or be taken to the next level?

This project started when we were in the middle of the pandemic, when we had a lockdown and there were still a lot of restrictions. We had to find a way to make the agile system work remotely during the pandemic, when everybody was based in a different location and most of the communication was taking place online. When the situation got better, it was a bit easier because everyone could come to the office and having a discussion was easier but before then, we had to find a way to do everything online. 

Apart from the changes we had to make in terms of how we work, we also had to make changes to how we manage our projects. Since this was a complex project, we had to organise many workshops and training sessions to ensure that our team had all the knowledge they required to be able to implement a product of this scale. We had mentors and advisors who provided continuous feedback and guidance. We also provided a lot of opportunities to the girls to try new things. For example, if somebody only had experience in  machine learning, they had a chance to try working on usability and see how they liked it. We wanted to keep it very open, to have an open environment for the girls to try new things, but at the same time to provide continuous mentorship and advice to make sure to guide them in whichever direction they want to go.

To introduce the topic of partnerships and liaisons, I would like to ask how the support and expertise from businesses came into play.

The good thing about this project is that we are focusing on agriculture and many sectors are looking for ways to empower and help the farmers and to promote agriculture. Since our product is an innovative idea, we garnered a lot of interest from various sectors. We are supported by the software sector, the banking sector and the agriculture sector, as well as the social sector. We have a software company in Nepal that is providing us the personnel and the consultation to create software that can be marketed. They have helped us in bridging the gap between the research and the market, so to speak. We have a bank that has been funding our project and they have also provided us reach to the network of farmers. We have collaborated with agricultural universities and agriculture-related companies who have provided us with consultations and data that we needed. And we have had a lot of support from different software developers and engineers who have been consulting us and providing us with direct feedback throughout the project. In a nutshell, we received a lot of help from different businesses and different centres of expertise.

Are there any particular ways in which these partnerships were created that you might find yourself commenting on?

We have been working for many, many years, doing this – doing different workshops, doing different internship programs – so we have expanded our network. Because of all the activities we have done in the past and how well we did them, we have secured our partners’ trust. There were a few partners that we were in direct touch with and with whom we have already collaborated in the past and they had connections. It just kind of went by the word of mouth from there.

We mostly covered the topic of continued implementation. Whether there are plans to expand would depend on the partnership and interest from other scientists and private partners, right? In that case, can you describe any opportunities or next steps that opened up, whether to the team, the organisation, the members, the project, during the duration of the project? I would be interested in hearing whether there was any unexpected attention or interest or connections that came about during the process, as well.

It was very obvious that there were a lot of sectors interested in contributing to agriculture and it was also very clear that we have a lot to do in agriculture. There were a lot of perspectives to explore and there were a lot of problems that required solving, and there is still a lot left to do in agriculture. We have received interest from different agricultural universities and researchers as well as funding agencies for collaboration. We have also received interest from different software and hardware companies who want to provide consultation. It was actually pretty nice; we were not expecting to get as much attention as we did. We are really grateful for all the help and support that we have received so far.

How will the experience feed into other activities or plans by Impacters?This also kind of touches on what we talked about earlier, with the difficulties in accessing the hardware. There is a part of our work that entails the stereotypes people have about the girls, but apart from that, we also have significant problems with the difficulty of access to hardware. Since realising this problem, we will continue looking into ways to allow young people to work on projects that make significant contributions to society. We want to offer opportunities to realise the projects in terms of procuring hardware, because that is not easy for individuals to do. We also want to make girls more confident when it comes to working with the hardware since it is even less common for girls to work with electrical components than it is to work on software. And we want to encourage the girls to try as many things out as they can and to get their hands dirty to realise the projects.

What kind of potential do the specific work and findings from this project have for different projects in relation to different problems and applications? For example, if it was applied to a completely different field, what kind of transferability would there be? Would the government or other organisations be able to take it on and how would they be able to transform it?

Again, the most important part to note here would be the open-sourcing of the project. We are going to open-source it and then government agencies, the scientific community, companies or individuals, whoever wants to use the results can basically pick up from what we have already done and build on top of it to avoid needlessly replicating the groundwork that we laid.

I was especially wondering what you think the most interesting or the most applicable parts of the groundwork would be for that kind of transferability. I am curious as to whether the relevant research would be the gathering of the datasets and working through how to make the data instruct the AI. Or would the more relevant part be mostly the software in and of itself rather than how it interacts with this specific data?

I think the most important thing to come out of this project for when others want to build on top of it would be the research that we did. In Nepal, a lot of people do not really want to do the research and development phase because it is a lot of work, it requires a lot of specialisation and it also has a lot of risk involved. The part that we have already done is the part that most people refrain from or are not interested in doing. Now, we have laid the groundwork for them and they can build on top of it. When it comes to relevant parts of the research, it is a bit of both. It includes realising what kind of data we need to fulfil the intended use. But it is also about what results you can get and how you can use those data points in service of that intended use, which makes the farmers’ lives easier.

Is there anything in particular you want to let people know about the project as a whole that is either particularly important or perhaps not immediately evident?

Yes; it relates to what I talked about before, about how a lot of people in Nepal don not focus on research and development. Instead, they prefer to buy products from abroad because it is much easier than investing in something uncertain, not knowing when or how a new product will be completed or if it will be worth the investment as a whole. But what we believe is that, without building our research and development infrastructures, we cannot create innovations that fit our requirements fully, which basically means that we are not truly able to help our society. We need to create resourceful spaces for young innovative minds, so that they are able to make meaningful contributions to society. The focus on research and development is something especially important that came out of this project.

Lastly, how can people follow and learn about the project and related materials [and] updates from Impacters and WODES?

We have social media for Impacters and for WODES: we have our Facebook pages and we have Instagram accounts so people can keep in touch and find all the updates about our programs on our social media. We also have websites for WODES and Impacters where they can get updates as well. We will be releasing papers and articles and blogs at a later point when we have conclusive answers. That will be worth keeping an eye out for as well.

Anitha Szabo is an intern at Ban Landmines Campaign Nepal (NCBL)/ Women Development Society (WODES) and a final-year undergraduate student completing a Honours Bachelor of Social Sciences in Conflict Studies and Human Rights at the University of Ottawa in Canada. In her university’s cooperative education program, she worked student jobs in internal communications, policy analysis and statistical survey testing.