
Sarvam AI is preparing to roll out a voice-enabled artificial intelligence app designed for India’s multilingual population, betting that consumer adoption, not just model development, will determine the country’s standing in the global AI race, co-founder and Chief Executive Officer Pratyush Kumar said in an interview with Business Today.
Speaking to Business Today Group editor Siddharth Zarabi, Kumar said that the upcoming application will allow users to interact with a locally trained large language model through text and speech across Indian languages, positioning it as a mass-market alternative to Western AI assistants.
“Once the app drops… you’ll be able to chat with it in English and Indian languages,” Kumar said. “You’ll be able to ask a question in voice… So general-purpose AI in your fingertips built in India with support for Indian languages.”
India’s AI ambitions increasingly hinge on translating sovereign models into widely used products. While domestic players have raced to train foundational models, adoption at population scale remains limited, a gap Sarvam aims to close with a consumer-facing platform tailored to local usage patterns.
Kumar said voice interaction is central to the strategy, reflecting both linguistic diversity and literacy constraints.
“India is a voice-first nation,” he said. “Typing Indian languages is not something that comes naturally… Voice is a natural medium.”
The app is expected to handle everyday queries ranging from education to agriculture and local search, illustrating a push toward utility rather than novelty.
“A student can ask questions on their topics… a farmer in the village can ask about some schemes… If you are just looking for the nearby restaurant, you can also ask that,” Kumar said.
Sarvam’s approach reflects a broader shift in India’s AI ecosystem from experimentation toward deployment, as policymakers and startups emphasise accessibility over frontier capabilities alone. The company trained its model using domestic compute resources with government support, part of a push to reduce reliance on foreign platforms.
“India should be building its own thing,” Kumar said, arguing the country is now in a “post-AI world” where large language models will underpin daily digital life.
Sarvam on 18 February launched two large language models, Sarvam-30B and Sarvam-105B, at the India AI Impact Summit. The company said both models will be released as open source, aiming to drive adoption among developers, enterprises and government agencies seeking alternatives to foreign AI systems. Since early February, the Bengaluru-based startup has introduced multiple AI tools for speech transcription, dubbing, translation across 22 Indian languages, optical character recognition and on-device AI.
Kumar warned that building the model is only the first step. The larger challenge is diffusion, putting the technology into everyday use across a country of more than 1.4 billion people.
“The transition from ‘I trained a model’ to ‘I’m using the model at population scale’ that is a problem worth solving,” he said.






