Tech giants are training their AI in Indian villages

In the town of Virudhunagar in southeastern India , a new field of work is developing around centuries-old temples. One of the town’s residents, Mohan Kumar, works as a data labeler in the field of artificial intelligence training. “My job is AI labeling,” Kumar told BBC television, “I collect data from different sources, label it, and train models so they can recognize and predict objects. Over time, they become semi-supervised and can make decisions on their own.” FROM BIG CITIES TO SMALL TOWNS For many years, India , with its major cities like Bangalore and Chennai, has been recognized as a global hub for IT services. However, in recent years, companies have begun to move towards smaller towns, where operating and personnel costs are lower. This trend, called “cloud farming,” has accelerated with the rise of AI-based businesses. Kumar stated that working outside the big cities does not limit his professional opportunities: “There is no difference in terms of profession. We work with the same clients in the US and Europe, whether in a small town or a metropolis. The education and skill level are exactly the same.” “WE BRING THE WORK TO WHERE THE PEOPLE LIVE” Kumar works at Desicrew, which was founded in 2005 and was one of the first companies to implement the cloud farming model in India. “We thought that instead of moving people to cities for work, we could bring the work to where the people live,” said Desicrew General Manager Mannivannan JK. “Opportunities have been concentrated in cities for a long time, and rural young people have been left behind. Our goal is to bring world-class careers closer to home and show that quality work can be done from anywhere.” The company offers outsourcing services such as software testing, data set creation for AI training, and content moderation. While artificial intelligence projects account for 30 to 40 percent of their business, Mannivannan said he expects this figure to “soon rise to between 75 and 100 percent.” Much of the work focuses on converting audio recordings to text (transcription). “Machines understand text much better,” Mannivannan said. “For AI to work naturally, machines need to learn to understand the differences in human speech patterns. That’s why transcription is such a crucial step,” he said. “GRADUATES’ POTENTIAL IS BEING WASTED” Bangalore-based NextWealth, founded in 2008, is also a pioneer of the rural outsourcing model. The company employs about 5,000 people at offices in 11 small towns across India, and 60 percent of its employees are women. Mythily Ramesh, co-founder and chief executive of NextWealth, told the BBC, “60 percent of university graduates in India come from small cities, yet IT companies generally only hire from metropolitan areas. This is squandering a huge potential.” Ramesh noted that most of her employees are from families with professions like farmers, weavers, or tailors, and that these families often have to pay for their children’s education with debt. The company, which expanded into artificial intelligence five years ago, receives approximately 70 percent of its projects from US-based clients. “The world’s most advanced algorithms are being trained and validated in small towns in India,” Ramesh said. “100 MILLION NEW JOBS COULD BE CREATED” Noting that artificial intelligence and generative AI have the potential to create approximately 100 million new jobs in the next three to five years in the fields of training, validation, and real-time computing, Ramesh added, “India’s small cities can be the backbone of this workforce.” Ramesh emphasized India's advantage over its competitors in this area, saying, "Countries like the Philippines can enter this space, but India's scale and early start give us a five- to seven-year lead. We need to capitalize on this opportunity." K. S. Viswanathan, a former advisor to the National Association of Software and Services Companies (NASSCOM), also noted that the growing AI outsourcing sector could reshape the country's digital economy. "Silicon Valley may be developing AI tools, but the day-to-day work that ensures those tools operate reliably is gradually leaving India's cloud farming industry," Viswanathan said. "If cloud farming grows at this rate, small cities in India could become the global hub for AI operations, just as they were for IT services two decades ago." However, Viswanathan noted that internet infrastructure and data security aren't at the same level in every region. "High-speed internet and secure data centers aren't always at the same level as in cities, making data protection a constant concern." According to Viswanathan, the real obstacle isn’t infrastructure, but perception: “The real challenge isn’t technical, it’s perceptual. International clients think small cities can’t meet data security standards. But the systems are robust. Trust is earned through delivery.” MODELS ARE BEING TRAINED WITH “FINE-TUNING” Dhanalakshmi Vijay, a NextWealth employee, is involved in a process called “fine-tuning.” She corrects instances where the model confuses similar objects, teaching it to distinguish, for example, a blue denim jacket from a navy shirt.