Main content

How Wefarm is using NLP and SMS powered by the cloud to help farmers share knowledge

Toni Sekinah Profile picture for user Toni Sekinah October 29, 2020
Built on AWS, Wefarm is a platform that connects farmers with other farmers, helping each other solve agricultural problems via the cloud.

Image of a farmer holding a mobile phone
(Image sourced via Wefarm)

There are an estimated 500 million smallholder farms around the world providing employment to 2 billion people. These agriculturalists know their land and crops inside out but the insight they have that could benefit others is not easy to share due to difficulties accessing the internet and geographic remoteness. Wefarm was set up to change this.

Founder and CEO Kenny Ewan was working with smallholder farmers in Latin America designing sustainable engineering projects in the early 2000s and saw the paternalistic top-down manner in which non-governmental organizations would attempt to disseminate advice and guidance to them. At the same time, he understood that farmers - who are just agricultural business people - have a wealth of knowledge about their crops, produce and land built up over generations.

The problem was that information was in siloes, so Ewan founded Wefarm as a platform to; connect farmers to each other for the purpose of knowledge sharing through SMS without the need for internet access. Initially a pilot project for smallholder farmer organization Producers Direct, the idea and service was developed and refined in Peru, Kenya and Tanzania. Adam Neilson, CTO of Wefarm explains how the service works:

A farmer, in a hypothetical situation, wakes up one morning and they discover a worm on their maize on their two-acre farm. They don't know what it is and their neighbors don't know what it is. They're 20 miles from the nearest village. What do they do? So Wefarm offers shortcodes that they can send a message to for free. That message is taken in by our service, we analyze that message for any content, any topics. We squeeze as much information out of that SMS as possible. And then we look in our database of about 2.6 million farmers for the most appropriate people that we think will respond to that question, with an accurate answer.

Wefarm is a communication and collaboration platform for small-scale farmers with roughly 65 staff and offices in four countries and two continents. The farmer-to-farmer digital network connects farmers' questions and answers both online and through SMS. Farmers fire off questions and answers to each other while Wefarm sits in the middle using machine learning to understand the request for information and patch it through to the right response. Neilson likens it to the way neurons and synapses work in the brain.

The questions run the gamut from ‘how do I get this unfamiliar crop to germinate?' to ‘how do I prevent soil erosion?' One farmer featured by Wefarm is Irene Katushabe who has a six-acre farm. From posting questions on the network she found out how to get her green pepper seeds to germinate, how to start growing ginger and how to improve her yield of tea.

The member farmers even get a star rating from the machine learning algorithm depending on how helpful other farmers find their expertise. With the knowledge gained through Wefarm, farmers have been able to get help identifying pests, diversifying their crops, sourcing seeds and better understanding pricing. Wefarm also has an online/offline marketplace where smallholder farmers can buy quality seeds.

The company first launched its service in February 2015 with a target of 1 million users. It now has 2.6 million users and has raised $21 million in investment. Though the headquarters are in London, it has offices across East Africa, in Kampala, Dar es Salaam and Nairobi.

The technology developed by the agriculture tech company recognizes the linguistic diversity of the countries it operates in. Its Natural Language Processing libraries have created models that identify Kiswahili, Luganda and Runyankore; regional languages that are spoken by 100 million, 8.5 million and 3.4 million people respectively. It also identifies English. The platform also makes allowances for typos, spelling differences and variations in dialect.

Pandemic impact

Since the onset of the Covid-19 pandemic and the huge uptick in information about it, as an organization dealing in knowledge, Wefarm has had to be wary of disseminating unsuspecting misinformation and malicious disinformation. Neilson said they attempt to counteract these types of messages. He says:

We have a responsibility to make sure that we're not disseminating false information and putting it into multiple hands. When we started seeing these types of topics coming in we were very conscious that we didn't want to make ourselves out as the organization can save the world. We wanted to be transparent about stuff. So we forward the message but we also send an automated message back to all the farmers giving them pointers, phone numbers and URLs that they could go to to get government sourced information, as well as, answering the questions as well.

Neilson describes the traffic profile in terms of volume of messages as "quite spiky" as questions often flood in during and after radio shows while farmers are listening in their fields. For several reasons, Wefarm was built on AWS and so is able to accommodate automatic scale-ups in growth to cope with incoming traffic.

There are three key reasons Wefarm was built on AWS according to the CTO. AWS was the legacy cloud provider at Wefarm when he joined in 2016. CEO and founder Ewan had brought in a developer who set up the MVP on AWS and Neilson said that it was easier to follow along the same path. Neilson was also more familiar with the products and with Amazon's set up as he had been using Amazon Cloud Services since 2010. The third reason for choosing AWS is the number of products it offers; the portfolio now includes 175 different products. Neilson explained:

There are so many parts to AWS now, there's so much on offer. It's kind of like a buffet of products and services that you can use and there are so many that are well suited to the project that we're working on. Probably everything from data science through to network routing, basic cloud server infrastructure, reporting, BI, monitoring. This massive array of various services was really attractive.

The company uses both public and private cloud with the public used primarily for admin tools and a VPC for internal settings. Wefarm benefits from open source software, according to Neilson and so the tech stack includes Kubernetes container-orchestration system, JVM and Clojure programming language which he describes as very capable despite not being mainstream and able to cope with Wefarm's applications. Neilson added Wefarm has a lot of interoperability with Java applications. Previously the company was hosting its own Kubernetes cluster on EC2 but moved over to Amazon's EKS. The reason for this was to eliminate the hassle of patching as new versions of Kubernetes are constantly released.

A different model

The fact that Wefarm is a for-profit business with a social mission is an aspect of the business that Neilson is keen to underscore. He says:

It's really important to highlight that you can run a for-profit business that has a social mission. From my perspective, there's not enough of us. I would be really keen to encourage other businesses to think about what mission they can have an impact on, but make sure it is profitable. Our job is essentially to try to 10x the farmers' businesses rather than our own business, and help them become flourishing organizations themselves.

While other farmer-focused agritech businesses aim to facilitate trade between urban and rural areas or on the dissemination of products and services from the Global North to the Global South, Wefarm's point of differentiation is its networked structure. The founders made a conscious choice to avoid the traditional set up of an NGO which, though well-meaning, can act in a paternalistic and disempowering way to farmers who as Neilson rightly says are just business people. Neilson says:

So, we're in the middle row, essentially our only responsibility is building some funky, machine learning models to understand what people are asking us, and then being able to forward that to people that we believe will give a response.

[Updated 10th November to correct total number of users and also added link to Jon Reed's story about WeFarm back in 2015].

A grey colored placeholder image