Log In

Infosys cofounder Nandan Nilekani commits second multi-year grant to AI4Bharat; total Rs 70 crore - The Economic Times

Published 1 day ago3 minute read
Business NewsTechAIInfosys cofounder Nandan Nilekani commits second multi-year grant to AI4Bharat; total Rs 70 crore
Infosys cofounder Nandan Nilekani commits second multi-year grant to AI4Bharat; total Rs 70 crore
ETtech
Nandan Nilekani – Infosys Co-founderPTI
Building on the success of his first grant, Infosys co-founder Nandan Nilekani has committed a second multi-year grant to AI4Bharat—an open-source initiative at IIT Madras that is building foundational AI models for Indian languages -- taking the total funding to Rs 70 crore.AI4Bharat aims to gather 15,000 hours of transcribed data from over 400 districts encompassing all 22 scheduled languages of India. In parallel, its in-house team of over 100 translators is creating a parallel corpus with 2.2 million translation pairs across 22 languages.

"This grant, issued directly by Nandan Nilekani, reflects his belief that AI4Bharat is one of the most impactful technology infrastructure projects in India. Their work will play a vital role in making India the AI use case capital of the world," a statement from EkStep Foundation, founded by Nilekani, said.

“AI4Bharat is building the infrastructure that ensures every Indian can access digital services in the language they speak. I strongly believe that India can be the use case capital of AI in the world. With our DPI foundation, we can build better AI, and in turn, AI can turbocharge DPI. Our belief is that AI should be inclusive, not extractive. It should amplify every human being’s potential. That’s our vision of AI for the people—AI to make lives better, AI to amplify human potential,” Nilekani said in the statement.

EkStep did not respond to queries from ET on how much the second grant alone is and its duration. Also, details of the quantum of the first grant is not known. In 2020, EkStep and AI4Bharat began investing in language digitisation, data infrastructure, and early models tailored to Indian linguistic diversity.

ET logo

"Applications were built within weeks of ChatGPT's release in 2022, ranging from chatbots in regional languages to integrations in governance and education. AI4Bharat’s models became the underlying infrastructure for this rapid deployment," the statement said.

AI4Bharat has become the language backbone of India’s AI stack, EkStep said. Its open-source speech, translation, and text-to-speech models now power key public digital platforms. Bhashini, India’s national language platform under the IndiaAI Mission, has adopted AI4Bharat’s models to scale multilingual services across governance, healthcare, and finance.The Supreme Court’s SUVAS system translates judgments into regional languages. National Payments Corporation of India has enabled voice-based Unified Payments Interface transactions in native tongues. Agricultural chatbots like Kisan e-Mitra use these models to deliver real-time information to farmers.

AI4Bharat’s language datasets—collected in all 22 constitutionally recognised Indian languages—have been released as public goods, openly available for use through the AI4Bharat website and integrated into AIKosh, India’s open AI repository.

Bengaluru-based Sarvam AI is the first startup to be selected to build an indigenous foundation model under the Rs 10,000 crore IndiaAI Mission, the government had announced.

"We are collaborating with AI4Bharat, in Indian language AI research, to build these models," Sarvam cofounder Pratyush Kumar had said.

“With the growing need for Bharat-specific AI models, and also reducing the possibility of a potential AI divide, the vision of 'AI for All' is extremely relevant for our country,” said Prof V Kamakoti, director of IIT Madras, in the statement.

The first grant given in 2022 established the Nilekani Centre at AI4Bharat, supported by EkStep Foundation.

Read More News on

Read More News on

Stories you might be interested in

Origin:
publisher logo
Economic Times
Loading...
Loading...
Loading...

You may also like...