In the busy streets of Mumbai, delivery driver Vineet Sawant spends his day weaving through traffic on his scooter. His job is fast packed, but in the beginning, he had another challenge - language.
Marathi is his first language, and he speak only a little English. Reading delivery instruction in English often slowed him down. Sometimes he made mistakes. “I used to ask other drivers for help to understand what to do,” he says.
Sawant works for Zepto, an online grocery delivery company that promises lightning fast service. When drivers struggle with instructions, every minute counts.
A year ago, Zepto partnered with Reveria Language Technologies to add a translation tool to its driver app. Now, Sawant can choose Marathi and see all instructions in his own language. “If the customer writes ‘ring bell,’ I read it in Marahti. I don’t have to guess anymore,” he says.
India has 22 official languages and hundreds of dialects. Without technology that work in local languages millions risk being left out of the digital world. This affects not only delivery work but also education, healthcare, banking, and public services.
Professor Pushpak Bhattacharyya of IIT Mumbai explains that building Artificial Intelligence (A.I.) language tools is not easy. “We need high quality data for each language. For regional and tribal languages, the data doesn’t exist isn’t digitised,” he says.
To bridge the gap, the Indian government launched Bhashini in 2022. The project collects language data, builds translation tools, and develops AI models for all 22 official languages.
Today, Bhashini runs 350 language models that have completed over a billion translation tasks. More than 50 central department and 25 state government use the technology. It powers multi language chatbots, translates government schemes, and make official information available in local languages.
Amitabh Nag, CEO of Digital India’s Bhashini Division, says the goal is simple, “We want rural users to access government services, financial tools, and information in their own language by voice or text within the next few years."
At IIT Mumbai, researchers are testing an AI tool to help people quit smoking, the program will talk to users in their preferred language, adjusting its tone and advice based on their stage in the quitting process.
Associate Professor Kshitij Jadhav says it will be highly personal. “It won’t be a generic script. It will respond with empathy, in the language the person is most comfortable with,” he explains.
For Sawant, the translation features has been life changing. He now delivers around 30 parcels a day, three time more than before.
“When the app speaks our language, we feel confident. We work better,” he says with a smile.