What is Sarvam AI? How founders Pratyush Kumar and Vivek Raghavan are building 'Bharat's' AI?

Sarvam AI has emerged as a frontrunner in India’s mission to achieve strategic autonomy in artificial intelligence. As one of the twelve indigenous companies selected under the ₹10,372 crore IndiaAI Mission, Sarvam is dedicated to building full-stack solutions tailored for the Indian context. Their vision is to move beyond English-centric models, creating a digital ecosystem that understands the vast linguistic and cultural diversity of the nation, from rural villages to urban centers.

The foundation of Sarvam AI rests on the expertise of its co-founders, Dr. Vivek Raghavan and Dr. Pratyush Kumar. Dr. Raghavan brings years of experience in building India’s digital public infrastructure, including his significant role at UIDAI (Aadhaar). Dr. Kumar, a former researcher at Microsoft and IBM, is a pioneer in open-source Indic AI through his work with AI4Bharat. Together, they are bridging the gap between cutting-edge research and population-scale deployment.

The showstopper at the India AI Impact Expo 2026 was the unveiling of "Sarvam Kaze," an indigenous AI-powered smart glass. Unlike traditional devices that act as mere screens, Sarvam Kaze listens, understands, and captures the user's environment in real-time. Prime Minister Narendra Modi personally debuted the wearable during the expo, testing its ability to respond to visual and auditory cues hands-free, marking a "sci-fi" milestone for Indian hardware.

Sarvam AI showcased a diverse range of platforms designed to solve real-world Indian challenges across sectors. Key highlights include "Saaras" for speech-to-text, "Bulbul" for natural-sounding text-to-speech, and "Sarvam Vision" for processing complex, multi-script Indian documents. These tools are supplemented by "Sarvam Dub" for multilingual content and "Samvaad," a platform for creating conversational AI agents that feel familiar rather than foreign to the end user.

At the heart of their ecosystem are foundational models like Sarvam 30B and Sarvam 105B, which are trained on trillions of India-specific data tokens. These Large Language Models (LLMs) are optimized for code-switching—handling the natural blend of English and regional languages known as "Hinglish." By utilizing the national compute facility’s 38,000 GPUs, Sarvam is delivering high-performance AI at a fraction of the cost of global competitors.

The "Arya" suite represents Sarvam’s foray into enterprise productivity, offering specialized tools for work, content creation, and real-time processing. To address connectivity and privacy concerns, "Sarvam Edge" allows AI models to run locally on devices like smartphones and laptops. This ensure that AI responses are instant, private, and functional even without an internet connection, making it ideal for the diverse infrastructure of the Indian landscape.

The success of Sarvam AI is deeply intertwined with the IndiaAI Mission’s broader goals of fostering talent and infrastructure. By addressing India-centric challenges, Sarvam AI may not only be creating a sovereign tech stack but also might generate a massive economic opportunities for the aspiring youth of the country.