Sushant Sachdeva urged India to focus on AI models for local languages rather than competing with the US and China on large foundation models. India should prioritise building artificial intelligence models for Indian languages instead of competing head-on with the US and China in developing large foundation models, said Sushant Sachdeva, a researcher at OpenAI and an academic at the University of Toronto.
“India must decide which stage of the AI pipeline it wants to fund. Competing on foundation models is extremely capital-intensive, and you’re up against companies building the most advanced systems in the world,” Sachdeva told ET in an interview. He was speaking in Bengaluru on the sidelines of the ceremony for the Infosys Prize 2025, where he received the award for engineering and computer science on Saturday.
Sachdeva said India’s biggest advantage lies in its linguistic diversity and population scale. “The number of people who speak Hindi or other Indian languages is far greater than those speaking many foreign languages, yet they are not adequately served by current AI systems,” he pointed out.
According to Sachdeva, India has both the talent and the technical capability to build strong local language models without developing foundation models from scratch. “You don’t need to compete with OpenAI or Alibaba. You can build on top of existing public models and publicly available data to deliver real value,” he added.
His comments come as India steps up efforts to position itself in the global AI race. Last week, Prime Minister Narendra Modi stressed the need for indigenous AI models that promote local content and regional languages during an interaction with select Indian AI startups ahead of the AI Impact Summit, 2026.
Several startups in the country — such as Sarvam AI, Krutrim, Gnani.ai, Fractal Analytics, etc. — are building foundational models and being supported under the government’s IndiaAI Mission.
Sachdeva was awarded by the Infosys Science Foundation for his contributions to algorithms and theoretical computer science. His research focusses on problems such as optimising transportation networks using geographical data, and improving the efficiency of data transfer across locations.
He is among six researchers under the age of 40 recognised this year across six disciplines, including economics, life sciences, and physical sciences.
Leave a Reply