In a post on X, the company said it has spent the past year quietly building full-stack AI for `problems of national consequence and complex enterprises,’ and is now expanding the effort into a dedicated vertical.
“On-prem deployments in air-gapped environments. Multi-modal data ingestion. Production-grade agentic workflows for institutions where failure isn’t an option,” a post by the AI startup on X said.
Sarvam is targeting customers that cannot rely on public cloud infrastructure or consumer AI tools, including regulated enterprises and strategic-sector institutions.
The company said the systems built by this team will have “dual use,” spanning enterprise and strategic sectors.
Sarvam introduced its first homegrown AI models — Sarvam-30B and Sarvam-105 B — at the India AI Impact Summit this year. These models support 22 Indian languages.
The company also rolled out several products, including Sarvam Vision for OCR and multimodal use, Sarvam Dub for translation and dubbing, and the Indus beta app for mobile and web. It also showcased Sarvam Kaze, its AI-powered smartglasses.
ET reported on March 25 that the foundational model startup is in early talks to raise $250-300 million from investors, including Nvidia, HCLTech, and Accel.
Sarvam’s cofounder Vivek Raghavan has said its monetisation strategy is centred on enterprise and government deployments. “We are primarily working with enterprises. Our approach is largely B2B. We’re also working with governments and different sectors,” Raghavan told ET earlier.
