AI sovereignty will happen, and India will achieve frugal AI, says IITM director Kamakoti
Prof V Kamakoti believes India is charting its own course to achieve authority in the field of artificial intelligence, and will come up with domain specific language models, which can be trained with fewer resources compared to LLMs.
Swetha Kannan
Wednesday January 07, 2026 , 3 min Read
Professor V Kamakoti, Director of Indian Institute of Technology Madras, is not unduly worried about India’s perceived lack of sovereignty in AI, and believes the country will come out with relevant domain specific models in due course of time.
“We have a lot of work going on in AI—fundamental work that builds AI systems, a lot of users of AI, a lot of papers from India are published in AI… We also have a lot of domain specific models coming up,” said Kamakoti, in an interview to YourStory.
Citing examples to illustrate the kind of work India has been doing in this emerging technology, he said, “BharatGen is there, there is AI4Bharat. Sarvam is doing extremely well... Perplexity is also our kid. So, the perception that AI is not there in India—I do not agree with it.”
BharatGen is India’s first government-funded, indigenous multimodal large language model (LLM) initiative. Sarvam AI is a startup building full-stack generative AI models and platforms specifically for India’s diverse languages and challenges, while AI4Bharat is a research lab at IIT Madras developing open-source datasets, tools, models and applications for Indian languages.
Incidentally, AI-powered answer engine Perplexity was co-founded by Aravind Srinivas, an alumnus of IIT-M.
Professor V Kamakoti, Director of IIT Madras, believes India will achieve sovereign AI and come up with its own models, which must be used carefully.
Despite having significant strengths in the form of talent and AI adoption, the general perception is that India lacks AI sovereignty compared to China and the United States. But Kamakoti is not perturbed by this.
“Sovereignty will come. We will have our own models, and we should use them carefully,” he reiterated.
Moreover, even the existing large language models (LLMs) have all been built on publicly available data from the internet, said the director of one of India’s most premier educational institutes.
“Whatever has been published and is available on the public domain has been used to develop models. So, every (person of the) public has a stake in it because they have contributed to that knowledge.”
On the need for more LLMs, he said, “I don’t know why you should be jack of all trades. It is not necessary.”
Kamakoti believes India is charting its own course to achieve authority in the field of artificial intelligence. He says the country will achieve success in creating frugal AI—born out of necessity due to scarcity of resources.
“Large-scale GPUs (graphics processing unit) are not available. There is an AI divide. These non-availabilities actually make us start working on frugal AI. This necessity is the mother of invention, where we start looking at how to do AI in a very cheap, very effective, but very light way.
“We are basically looking at domain specific language models. This model may not require that (large) level of GPU and can be trained with lesser things. Once that’s trained… India will have a lot of domain specific language models which will be sovereign.”