New Delhi: Artificial intelligence is entering a phase where innovation is accelerating rapidly and reshaping everyday digital experiences, according to Pratyush Kumar, CEO of Sarvam AI.
As India hosts one of the major events of this year, the IndiaAI Summit, Kumar emphasised that the country has a significant role to play in ensuring that technology benefits everyone, not just a select few. He said AI is not something ordinary people fear, but rather a tool that is empowering individuals to become creators – something many would not have imagined even five years ago.
“We are in the midst of an extraordinary technological revolution, where innovation is accelerating at an unprecedented pace and multiple transformations are unfolding simultaneously. This moment is a celebration of that impact. I genuinely believe India has a significant contribution to make to the world, one that ensures technology works for everyone, not just a few. If you ask an ordinary person whether they are afraid of AI, most would likely say no. I sense that artificial intelligence is empowering people. The AI of today is enabling individuals to become creators, something many of them would not even have imagined for themselves five years ago,” he stated.
Ahead of the IndiaAI Impact Summit, Sarvam AI announced the launch of Sarvam Edge, an on-device artificial intelligence platform designed to run directly on smartphones and laptops without relying on cloud servers or internet access. It was said that the platform is designed to make AI faster, more private, and usable across everyday consumer hardware.
Kumar said: “Sarvam Edge is our dedicated effort to bring intelligence to run offline and on-device. Our goal is to make AI that is efficient, private, and accessible everywhere…We are showcasing our speech recognition, speech synthesis, translation and document digitisation models running on-device. This comes on the back of our research efforts to make models super small in memory and compute footprints, while being close to the accuracy of much larger models. We believe running models at the edge will truly unlock population-scale AI in India. And we are working with partners to make this a reality.”
Sarvam Edge shifts AI inference from the cloud to the device, leveraging the computing power already available in modern smartphones and laptops. By operating locally, it removes network delays, improves reliability, and ensures that user data remains entirely offline. The company said this approach strengthens data privacy, eliminates per-query cloud costs, and maintains consistent performance even in regions with weak or no connectivity. The platform integrates multilingual speech recognition, speech synthesis, and translation models optimised for edge deployment. Its speech recognition model supports 10 major Indian languages with automatic language detection, allowing users to speak naturally without manually selecting a language.
The text-to-speech system works across the same languages using a compact unified model to deliver consistent voice output. The translation engine supports 11 languages, including English, enabling bidirectional translation across 110 language pairs, all without requiring cloud connectivity.
Sarvam AI said the models are engineered to balance accuracy with low memory consumption and fast response times. Real-time transcription processes speech faster than live audio, while translation provides near-instant responses suitable for interactive applications. The models are optimised for existing mobile processors and laptops, enabling full offline AI functionality without specialised hardware or excessive storage requirements.
Sarvam Edge adds to the company’s expanding product portfolio, which includes Sarvam Studio, Saaras V3, Sarvam Arya, Sarvam Audio, and Sarvam Dub. With this launch, Sarvam AI is positioning on-device intelligence as a core element of future digital infrastructure, aiming to expand AI adoption across education, accessibility, finance, and everyday productivity while reducing dependence on constant cloud connectivity.
