People of ACM - Vivek Seshadri

April 21, 2026

How did you initially become interested in developing a digital work platform?

In early 2017, when I was exploring a variety of ideas in the intersection of technology and development, we noticed some clear trends. Technology was getting better and cheaper. Smartphone penetration was increasing and internet data costs were reducing, particularly in India. Yet the majority of new digital users (often from low-income households) were using their smartphones primarily for entertainment, social media, and messaging. We asked ourselves, "If people had access to a connected device, can we provide them with productive work that can pay them supplemental income and potentially upskill them?”

Almost around the same time, there was a growing need for language data for training new multilingual models. This need for language data matches the skillsets of our target communities. However, there was no good smartphone-first digital work platform that could connect our communities to language data work. This was the genesis for our research around this topic and, ultimately, the development of the Karya digital work platform.

A stated goal of Karya is to create high-quality data sets by solving challenges that exist at various levels of the data ecosystem. How is Karya’s approach an advancement over previous methods?

When we started this work, digital work platforms were mainly accessed via a computer with internet connection. Unfortunately, most people in remote rural communities did not have access to this setup. We built our platform to cater to the constraints of our communities. Our platform is mobile first. We ensure it can continue to run on old android versions, is accessible (all instructions and content are in the local language of the speaker) and inclusive (it can be set up to run even in regions without internet connectivity).

We also ensured that our platform had a clear priority for worker well-being. Our workers are paid high wages (often much higher than local living wage rates) and we ensure that we never provide tasks that can harm the mental health of our workers.

Our approach has enabled us to reach users in remote communities, in turn enabling us to collect diverse datasets.

What were the engineering challenges in developing this platform? As a social entrepreneur, why do you think Karya will be economically viable in the long term?

On the engineering side, the main challenges were reaching and supporting workers with limited connectivity and devices, designing tasks and interfaces that work in many languages and literacy contexts, and building pipelines that scale to millions of microtasks while keeping quality high and latency low. We had to make the platform usable on basic smartphones and in low-bandwidth settings, and to support many local languages and task types.

On viability: the demand for high-quality, ethically sourced AI data is growing, and buyers are increasingly willing to pay for provenance and impact. Karya’s model turns fair wages and worker welfare into a quality and trust advantage rather than a cost to minimize. As long as that demand exists and we keep delivering both quality and impact, the unit economics can support the social mission and make the platform sustainable.

How do you see this approach expanding to enable upward socio-economic mobility for more people?

Our research shows that even small engagements with flexible digital work can have dramatic positive effects, particularly for women in rural communities. People report improved confidence in using a smartphone, greater comfort in handling money, and improved dignity and household standing. While our main goal is to increase the amount of income we enable per individual household, we are expanding our work to more regions and languages in India as well as other countries in the world. We are also exploring ways of making the above factors measurable as primary effects rather than secondary side effects.

In broad terms, how will digital work influence the development of AI technologies in the coming years?

Modern AI technologies significantly improve efficiency and access to information, services, and opportunities. Natural language interfaces to AI (text/voice) have dramatically improved who has access to AI. However, a large number of languages and dialects are still left out of the process or deprioritized. The best way to bridge this gap is by dramatically increasing the amount of data available in all languages. Digital work is key to building these language datasets. In addition to enabling more equitable AI technologies, it also infuses income into marginalized communities that can benefit from those technologies.