Karya, a non-profit launched in 2021 in Bengaluru, founded by Manu Chopra and Vivek Seshadri calls itself ‘the world’s first ethical data company’. It sells datasets that it collects from rural India to big tech players. Unlike a for-profit entity, Karya uses the money earned from selling the data to simply cover its costs and uses the remainder to support the rural poor in India. By partnering with and providing jobs to rural India, Karya is able to provide them with de-facto ownership of the data that they provide. Whenever this data is sold, the providers (people categorically below the poverty line) receive a generous compensation that is over and above the minimum wage that they anyways receive. It’s a model that doesn’t exist anywhere else in this industry.
Storyboard18 caught up with Nitin Jerath, head of design at Karya to understand how this model works, how Karya is able to navigate the rural market efficiently, provide high paying long-term jobs and how they are effectively using technology to help bring people out of poverty.
Edited excerpts.
What does Karya do ad what is the goal?
Our goal is to provide a means of working to those who are underrepresented and who don’t get opportunities for work otherwise. We work with a lot of rural communities that fall below the poverty line and who do not have access to work or high paying work. So Karya’s goal is to provide dignified digital work to people in rural India. We currently create datasets for tech companies. If a company wants a dataset, we provide it to them. We want to build an ethical data ecosystem which currently has a very bad rep for its unethical practices.
How are you executing this currently?
How we do that currently is through our mobile platform, which is our smartphone app, which anyone can download and use in their native language.
What are the practices that you are undertaking to achieve this?
‘Karya Earn’
There are three programs that we have in the pipeline. One is called ‘Karya Earn’. It is essentially what I briefly touched upon. We get high paying data work and we make it accessible to these people. So as you know, the data industry is a booming industry and all these AI (artificial intelligence) and ML (machine learning) models required data sets for them to train on.
In a lot of cases, the means of creating these datasets has not been very ethical, if you look historically. So Karya comes in there and sort of crowd sources this data. And since we come in from an impact angle, we’re an impact first company, we divert the majority of the wages directly to the people. These people also retain ownership of this data. So what we do is we license these datasets to our clients under something called KPL, which is Karya Public License, which one of our co-founders spent the majority of the year working on. What this essentially means is that any time this data is sold again, these people would get paid again. Say today I’m building a Marathi dataset for Microsoft. That dataset can be sold to Google or Facebook or whoever may be interested in building tech in that language or may find that dataset useful. And that would fetch royalties to people who contributed towards making it in the first place. The goal through this to maximise earning opportunities for our communities.
‘Karya Learn’
Since our primary focus is dealing with poverty, we recognize that it’s a complicated problem. You can just give someone money and expect them to build their own pathway out of poverty. So what we’re piloting this year and even did a study with 96 people for is called ‘Karya Learn’. What this means is that people can do work and by doing the work, they can also gain critical skills and knowledge.
For example, let’s say that Apple wants Siri to talk in a regional language – Kannada. For that to happen, they would need to train Siri on hundreds of hours of data of actual people talking in Kannada. So the model or the voice assistant in this case can understand how different people pronounce different words in different ways and what the natural way of speaking the language is.
In this case, they would reach out to someone like Karya with a requirement of 200 hours a day of Kannada. We would break this huge chunk work down into micro tasks. These micro task can be accessed by our communities through our app. By doing these tasks, they can contribute to our data sets and earn significantly more than what they would earn doing anything else. And they could do this sitting at home, it’s highly flexible.
We’re able to handle this in a very nimble fashion. We have executed projects completely remotely as well, which brings down operations costs and stuff like that. This helps us maximize worker income.
Karya Learn would basically take a task like this where people are sitting and reading out sentences and make those sentences educational. The client normally doesn’t have any requirements about what people are reading out. It just has to be diverse and unique. So what is being read is also being learnt. And they get paid to do it. It’s interesting, in a way you get paid to learn.
Through this model, we plan to build modules and partner with people who already have modules and then convert it through this mechanism to reach hundreds and thousands of people who can essentially read out sentences, create datasets that contribute to tech and also learn in the process. That is like a natural progression of Karya Learn. Anyone who comes in to work with us would ideally have access to learning modules to take up.
‘Karya Grow’
Finally, the third step would be Karya Grow, which we have planned for next year. Through this we are looking at building career pathways. We don’t see our current Karya work as our community’s full time job. It’s a means to earn a critical supplementary income. Thus, enabling people to have better access to education, better amenities because they have more money to spend. Through ‘Grow’ we would help them build career pathways that are more longterm and eventually bring entire families and communities out of poverty.
What are some of the insights that you’ve received from your communities? And what are the learnings that you have from your work?
This whole framework of Earn, Learn and Grow has been put together because of feedback that we’ve received from communities and through our constant interactions with them. Because as you can imagine, this is a very ops-heavy process since we have to interact with a lot of communities on ground. Somewhere in that cycle we realized that just data work will not be enough. We will need to expand the scope to learn and grow and long term career opportunities eventually.
We also learnt that people are very willing to work and a lot of these people haven’t worked before. Think 60% or more. In a lot of cases, more than 60% of our workers are women. And that too women who haven’t worked ever. Interestingly, they end up working more efficiently than those who work in a BPO model.
We learnt that if you pay people well, if you pay people the fair wage, the data quality also seems to improve. The fact that people are exploited for technology is a market failure. You’re selling datasets sometimes 200x of what you’re paying your workers. That what Karya is focusing to solve.
What sort of impact this guy they are currently having and what impact do you wish for it to have?
We have over 200 partners on ground so that we can execute projects across India. We’re building that capacity to get in more and more work for the rural communities. Part of that is internal capacity building and that’s something that we currently have been working on. We realise the value of this data and the languages that people speak. Your native language may not be documented or does not have a dataset. Indian languages are highly underrepresented in the dataset market. Thus, creating high quality datasets for Indian languages to be used by tech giants is one thing that we’ve been doing well.
What we’d like to do is basically build pathways for people out of poverty. We’d like to see people go through our Earn, Learn and Grow pathways and benefit from them at scale.
We are currently able to pay our workers 20 times the minimum wage. This is done through crowdsourcing the Karya Public License that helps pay these workers royalties.
Who exactly is your target group and how do you reach out to them?
Our target group is basically anyone and everyone who would fall below the poverty line. Those who can read and speak their local language. That’s the only criteria you need to work with Karya.
Currently, we go through partners – NGOs. We have over 200 partners across India. They’ve done their due diligence already. So they work with communities who genuinely need help the most. Our focus currently is in identifying who these top partners are to work with. These partners are people who work with communities on ground at a grassroots level.
Tech savviness and availability still seems to be an issue in rural India. Since, most of Karya’s work initiates through its mobile app, how are you able to navigate this?
We’re working pan-India which means different communities, different languages, different understanding of even visuals. For example, Apple did a very good job initially when they designed iOS. Swipe to unlock and all of those usage patterns that they built were taking inspiration from real world things. In design terms it’s called neumorphism.
However, if we were to use this concept from the west, it wouldn’t work for us. If we used an icon of a house for something, to rural India, it may not look like a house. We try and build the app in a way where anyone with a smartphone can navigate it.
We’ve introduced audio prompts. The phone can actually guide you if you try to sign up. The way in which they interact with their task. It’s drawn inspiration from apps that they may use on a very regular basis, like WhatsApp. We try and derive all these usage patterns from stuff that they may actually already use. Of course, if someone hasn’t used a smartphone before, they require some amount of training. But we’ve also kept that in mind when we’re designing these interactions so they don’t have to learn more than they need to. If they know their local language, they can just read and know what to do next and just be able to do it.