Manu Chopra, the co-founder of AI startup Karya, will speak at NDTV AI Summit 2026 on the theme "Should AI be Sovereign?" at ITC Maurya, New Delhi. The summit brings together people across the globe, including tech leaders, policymakers and entrepreneurs.
A Stanford graduate, Chopra launched Karya in 2021 with an aim to make artificial intelligence work for everyone and not just English speakers. He wanted AI tools to understand local languages so that more people could benefit from the technology.
Who is Manu Chopra?
Manu Chopra began working as an instructor at Stanford University in 2015, where he learned web technologies by building real-world projects focused on social impact. In 2016, he joined Microsoft as a research intern, where he worked on solving global challenges in healthcare and improving low-cost access to technology.
After taking up a few smaller projects, he returned to Microsoft in 2018 as a research fellow. There, he focused on tackling extreme poverty by creating opportunities that gave the poorest communities access to dignified work. In 2022, he co-founded Karya.
Why does Karya pay $5 an hour?
The company pays rural, low-income Indians up to 20 times the local minimum wage to help build AI systems, according to Forbes.
Karya pays $5 (approx. Rs 455) per hour to people for speaking in their own languages. People are asked to read simple sentences aloud in their mother tongue, which helps the company create voice recordings and language data. This is later used by AI systems to understand Indian languages and accents better.
Once AI is trained, people are then shown images and asked to describe them in their own language. This helps AI learn how people naturally speak and explain things. The next step is subtitling and advanced tasks. At this stage, Karya pays $10 an hour (approx. Rs 910) to advance train their models.
Some of the more advanced and technical tasks at Karya pay even more - up to $25 an hour (approx. Rs 2,270). These usually involve complex language work that requires higher accuracy and skill.
Karya helps locals earn better
Swarnalata Nayak, an artisan from Odisha, struggled to support her family despite working alongside her husband. Their income was so low that at times they earned just Rs 1,000 a month. However, she revealed that things changed when her friend introduced her to the Karya Project.
Through Karya, she simply records her voice in Odiya on her phone for AI language data. Within just one week, she earned Rs 4,000, far more than what she had earlier made in a month, according to The Better India.
How does Karya help AI startups?
Big companies such as Google or Microsoft need huge amounts of language data to train their AI models. For example, if they want their AI to understand Hindi, Tamil or Marathi better, they need thousands of hours of real speech in those languages.
That's where Karya comes in. It connects with rural Indians and asks them to record sentences, have conversations, describe images or do language-related tasks in their mother tongue. This speech and text data is then used to train AI systems.
"If you want to build inclusive AI, you have to employ the communities. The reason AI is worse off in Tamil than it is in English is that fewer Tamil-speaking people have trained AI models. And that's what we need to change," Chopra said.
