Amitabh Nag: Breaking Indian language barriers in AI

As the CEO of Bhashini, Nag is helping build the government's platform to make digital content and services more accessible across India's diverse languages

  • Published:
  • 06/06/2025 02:03 PM

Amitabh Nag, CEO, Bhashini Image: Amit verma 

Amitabh Nag’s journey to leading the Digital India Bhashini Division began with his experience in the tech industry. Before joining the Ministry of Electronics and Information Technology (Meity), he worked with tech giants, including Tata Consultancy Services and HP, gathering expertise over four decades. With this wealth of knowledge, Nag joined the Indian government to spearhead Bhashini, an initiative launched by Prime Minister Narendra Modi in July 2022.

Related Stories

  • AI unlocks new avenues for creativity, strategy, leadership: Salesforce's Srini Tallapragada
  • India is evolving from an AI adopter to AI leader. Here's how
  • India is metamorphosing from AI adopter to AI leader

The AI-powered language translation platform aims to make digital content and services more accessible across India’s diverse languages. It bridges the language gap with the use of artificial intelligence (AI) and natural language processing (NLP). Through Bhashini, Nag is building a suite of communication, translation, and transaction-related services.

“Apart from breaking language barriers, we’re also making digital content more accessible,” he says. “We’re working on improving voice-to-text capabilities, recognising printed text and handwriting, and creating more Indian language content online.” Bhashini supports over 22 Indian languages, including Bengali, Hindi, Tamil and Telugu.

The AI-based translation tool is designed to facilitate communication between people who speak different languages. As a crowdsourced platform, it allows individuals to contribute to teaching it new languages and dialects. This enables people to communicate in their own language, even when interacting with someone who speaks a different language. The platform handles around 8 to 10 million translation requests daily, which amount to about 300 million translations per month. “Cumulatively, we’ve handled over 2 billion translation requests,” Nag reveals.

Bhashini’s models were developed from scratch using encoder-decoder technologies (a type of AI architecture used for tasks like language translation and text summarisation) and other fundamental AI technologies. According to Nag, the development process involved collaboration with around 70 research institutes across India. “In 2021-2022, when AI was still in its early stages, we reached out to institutes with expertise in language, AI and voice technologies,” he says. They formed consortiums of eight institutes for eight projects, each tasked with building models from the ground up.

As Bhashini’s models take shape, the question of monetisation arises. Nag notes that the government will decide whether the technology will remain free or incur charges. “There are expenses from paying third-party vendors, and we’ll need to consider how to monetise it,” he says. Since the models were built using public expenses, the focus won’t be on recovering R&D costs but rather on covering operational expenses and potential improvements. The approach will depend on how people use the technology and the value they derive from it. The total mission cost is approximately ₹495 crore so far.

Also read: How seasoned IAS officer Abhishek Singh is leading India's AI mission

The platform’s offline capabilities and voice-based interfaces will ensure that even regions with limited internet connectivity can benefit from digital services, effectively narrowing the digital divide, explains Brijendra Kumar, partner and lead, Digital Government, KPMG India.

In agriculture, its deployment through the PM Kisan chatbot has facilitated farmers in accessing scheme-related information in their native languages. “Sectors like health care and education stand to gain significantly; for instance, integrating it into telemedicine could bridge language barriers between doctors and patients.

Similarly, multilingual educational tools aligned with the National Education Policy can democratise learning for students across linguistic backgrounds,” adds Kumar.

Nag explains that the next step is to further develop key products like document translation, video translation, and speech-to-speech translation. The goal is not just to build these products but to work closely with the ecosystem, especially state governments. “The idea is for these models to be co-owned by the respective state governments or communities,” Nag says. This collaborative approach, he says, will drive faster growth and improvement of the models.

As of March, Nag has taken on additional responsibilities in the India AI Mission. One of his key focuses is AIKosh, where he’ll be developing a repository of AI applications and identifying problem statements that can be addressed through AI.

Last Updated :

June 06, 25 02:56:32 PM IST