Sarvam AI, a Lightspeed-backed artificial intelligence startup, has announced a strategic partnership with Yotta, a prominent AI infrastructure provider, to roll out a comprehensive suite of generative AI tools designed for both enterprises and developers. This collaboration follows Sarvam AI’s successful $41 million Series A funding round in December 2023, led by Lightspeed Ventures, with support from Peak XV Partners and Khosla Ventures.
Generative AI Offerings
The partnership with Yotta enables Sarvam AI to launch several innovative generative AI products, including Sarvam Agents, Sarvam Models, and a developer-friendly API. These offerings are underpinned by the Sarvam 2B model, which is specifically trained on Yotta’s advanced infrastructure. Vivek Raghavan, co-founder of Sarvam AI, highlighted the advantages of utilizing Yotta’s data centers, which are equipped with NVIDIA H100 GPUs optimized for Indian enterprises. Raghavan emphasized that these GPUs offer a significant edge over other cloud providers like Azure and AWS, which currently lack similar capabilities in India.
Sarvam Agents and Indic LLM
Among the key products launched is Sarvam Agents, a voice-enabled, multilingual platform designed to serve as custom business agents. These agents are deployable across various channels, including telephone, WhatsApp, and in-app interfaces, supporting 10 Indian languages. Raghavan noted the cost-effectiveness of these AI-driven agents, which can significantly reduce customer service expenses for enterprises.
Sarvam AI has also introduced India’s first foundational, open-source 2B Indic Large Language Model (LLM), trained on a vast internal dataset of 4 trillion tokens. This model is engineered to efficiently represent and process 10 Indian languages, offering a locally-tailored alternative to larger, more general models.
Shuka 1.0: India’s First Open-Source AudioLM
Another groundbreaking product is Shuka 1.0, India’s first open-source AudioLM. This model, an extension of LLaMA 8B, supports Indian language voice input with text output, promising higher accuracy than existing frontier models. Sarvam’s offerings are poised to be transformative across various sectors, including BFSI, healthcare, legal tech, and more.
Partnership with Meta
In addition to the Yotta collaboration, Sarvam AI has partnered with Meta to create a specialized WhatsApp stack tailored for enterprise use. This stack includes features such as conversation history, business instructions, targeted marketing, and consumer feedback, with the added capability of enabling enterprises to accept payments directly via WhatsApp. This move capitalizes on WhatsApp’s extensive reach in India, with approximately 500 million users and 20 million monthly transactions.
Focus on Smaller, Task-Specific Models
Sarvam AI differentiates itself by focusing on smaller, task-specific AI models, as opposed to larger models like OpenAI’s GPT-4 and GPT-5. According to Raghavan, these smaller models are more cost-effective, exhibit lower latency, and are better suited for high-frequency use cases, making them a practical choice for enterprises.
Founded in July 2023 by Vivek Raghavan and Pratyush Kumar, Sarvam AI continues to push the boundaries of AI innovation in India. The startup, originating from AI4Bharat, a venture backed by Infosys co-founder Nandan Nilekani, is rapidly establishing itself as a key player in the generative AI space with its full-stack offerings.
For more details on Sarvam AI’s latest products and partnerships, visit Sarvam AI’s official website.
