The Institute consistently partners with government bodies and global non-profit organisations to help those who need it the most. Currently, they are building AI-based solutions particularly in the agriculture and healthcare domain, and extending their prowess to education. Over time, they have been successful in tackling challenges like pest management for cotton farms, maternal, newborn, and child health, tuberculosis and Covid-19.

As CEO of Wadhwani AI, Shekar Sivasubramanian is driving the organisation’s efforts to build AI-driven solutions and ecosystems for the benefit of millions across the developing world.

In a freewheeling chat with Jibu Elias — Content and Research Head of INDIAai, Sivasubramanian opens up about the journey of Wadhwani AI, the solutions they are building, and more.

The journey of Wadhwani AI

Sivasubramanian recalls the humble beginnings of the Institute way back in 2018, when their small team was more focused on research. Post-Covid, they believe they have gained more clarity, having rediscovered themselves.

Commenting on the nature of the AI space, Sivasubramanian explains how there exist academia and tech companies, a single unit that serves an urbanised, digitised and literate community. There’s also the market that represents the public space, government, and the poor population.

“The challenge was to find someone between these two spaces. Pure academia likes to solve problems, runs back to the lab, and is not interested in understanding ground level realities. We recognised a clear space that’s one of the most underserved and under-filled. It requires you to tackle a problem with a certain level of understanding and without any bias. That’s where we fit in,” adds Sivasubramanian.

Here, the agenda at Wadhwani AI is to work with large institutions, often the governments, who can take things to scale. The Institute’s mission has always been to make an impact on the underserved population.

When they started out, it was important for them to participate in the market extensively, understand the existing problems, recognize the challenges, and work on them.

“There’s a common rhetoric that you hear – that there’s no digital pipeline. We don’t get into that discussion at all. Instead, we say whatever is available right now is the basis for us to start working on problems. In trying to create the perfect environment, you lose several years. Plus, it’s always a work in progress,” Sivasubramanian shares.

A future-forward mindset

Adopting the approach they have taken doesn’t just require hard work, it also requires a mindset to create a set of solutions to become toolkit-ready. In this manner, they have made themselves adept at quickly giving demos on a range of problems that they’ve already addressed.

“We work closely with the government to identify spots where these things can be plugged into. You should have the ability to take semi-crafted solutions in AI and fit them into the ecosystem,” says Sivasubramanian, adding that their pest management solution is being developed to help reduce crop losses through an AI app.

The benefactor of their solution is a cotton farmer, who has installed pest traps in his farm and works closely with farmer welfare groups to manage pest infestations. The good news is that this app works on a simple smartphone and can even be accessed offline in places with low network coverage.

They started out small with 500-1000 farmers in a few states. In the recent Kharif season that ended in January, Wadhwani AI was successful in contacting about 35,000 farmers directly, some of whom don’t even have smartphones but are a part of the network.

“This way, you are able to impact about 1-1.5 lakh farmers out of 6 million farmers. Our aspiration in 2023 is to try and maximise our reach to these farmers,” he adds.

Addressing challenges, one step at a time

While there’s great hype and excitement around AI, Sivasubramanian believes the implementation requires only a few simple steps. Most importantly, it is about solving a problem as quickly as possible, largely because a small amount of data is accessible at a given point. By sitting in a lab and purifying efforts with a sanitised amount of data isn’t helpful in any way.

“But at the same time, it is important to sensitise the marketplace. There should also be human intervention to assist in the decision-making process so that if something goes wrong with AI, the human can keep iterating the model,” he mentions.

Wadhwani AI’s energies are currently focused on agriculture, education and health. They have also recently signed a pact with the National Skills Development Corporation. Some of the prominent solutions that the Institute is working on at this point includes a clinical decision support system and integrated disease management.

Sivasubramanian makes a valid point that agriculture is still in the process of maturing its digital imprint. Despite the challenge, the Institute has identified a few problems to work on. During the course of the year, Wadhwani AI is confident that things will progress in the right direction.

Before ending the chat, it became starkly clear that while building solutions has always been critical for Wadhwani AI, they are also advocates of responsible or ethical AI. They have explicitly addressed it by providing solutions that have no biases.

Even for a tool as extensive as ChatGPT, there are inherent biases simply because the corpus of knowledge tends to be linguistically, culturally or even religiously biased, educates Sivasubramanian

“For institutes like us, it is important to constantly monitor and implement, because our problems are very specific. We aren’t trying to provide a generic solution,” he concludes.