FE Leadership Awards | Emerging Entrepreneurs: Pratyush Kumar, Vivek Raghavan – AI that speaks India’s language – Industry News
For Vivek Raghavan and Pratyush Kumar, the challenge was never just to build another artificial intelligence model, but one that could understand India’s languages, documents and everyday contexts. When they founded Bengaluru-based Sarvam AI in 2023, the global AI race was already dominated by companies with deep pockets and vast computing resources. Their bet was that India needed its own AI capabilities rather than depending entirely on models built elsewhere.
The founders brought complementary strengths. Raghavan had spent years building technology businesses before working on Aadhaar’s technology infrastructure and improving AI for Indian languages. Kumar, a computer engineer with a PhD from ETH Zurich, had done AI research at IBM and Microsoft and was associated with AI4Bharat at IIT Madras, which develops language technologies for India.
Those backgrounds shaped Sarvam’s founding idea: foundational AI models that work across Indian languages and serve local needs in public services and businesses. It began with speech recognition, text-to-speech and document understanding, areas where linguistic diversity, mixed scripts and poor-quality records trip up general-purpose AI. Its models transcribe speech, generate natural-sounding voices and extract information from documents in Indian languages.
Sarvam has since grown into a broader AI platform spanning foundational models, computing infrastructure and applications for enterprises and government. The aim is to build the technology stack in India, from training to deployment, at a cost suited to a price-sensitive market.
That takes capital, computing power and specialist talent. In June 2026, Sarvam announced the first close of a $300 million Series B round, raising $234 million at a post-money valuation of $1.5 billion. HCLTech and Bessemer Venture Partners joined existing investors Khosla Ventures and Peak XV Partners. The money will support larger models, computing capacity, research and enterprise deployments. Sarvam is also moving into agentic AI, where systems carry out multi-step tasks, and coding models.
Commercial traction is emerging in banking, insurance, government services and document processing. Sarvam says its conversational platform handles more than two million interactions a day, and its document technology is digitising millions of pages. Such deployments offer a path to revenue beyond offering models to developers, but the company must prove its systems deliver reliable results at scale and compete on cost with global giants.
For Raghavan and Kumar, the next phase is turning the case for Indian-built AI into a sustainable business. That means sharper model performance, attracting specialised researchers and converting enterprise interest into repeatable deployments. Their larger goal is an AI ecosystem built and run in India that also serves needs beyond its borders.
Source: www.financialexpress.com