Sarvam AI is an Indian generative AI startup focused on developing large language models (LLMs) tailored for Indic languages. Here are the key aspects of what Sarvam AI does:

Core Offerings

Indian Voice LLMs: The company is building generative AI models specifically designed for Indian languages, starting with Hindi and aiming to expand to other languages. This includes creating a natural voice-based interface for LLMs, making AI more accessible in local contexts.

Generative AI Applications: Sarvam AI aims to facilitate the development and deployment of generative AI applications across various sectors such as education, finance, healthcare, and customer service. Their focus is on making these applications accurate, cost-effective, and robust.

Enterprise Solutions: The company is developing an enterprise-grade platform that allows businesses to build generative AI applications efficiently and effectively.

Collaborations

Partnership with Microsoft: Sarvam AI has collaborated with Microsoft to leverage Azure's infrastructure for training and deploying their Indic voice LLMs. This partnership aims to empower developers in India to create scalable generative AI applications quickly and responsibly.

Key Products

Sarvam Agents: Voice-enabled, multilingual agents that can be deployed via telephone, WhatsApp, or in-app.
Available in 10 Indic languages, these agents are designed for various business applications and are priced at ₹1 per minute.
Sarvam 2B: An open-source large language model (LLM) specifically for Indic languages, trained on an extensive dataset of 4 trillion tokens.
Capable of tasks like translation and summarization in vernacular languages.
Shuka 1.0: An open-source audio language model based on Meta AI’s LLaMA 8B model.
Supports voice input and text output for Indic languages, enhancing accuracy compared to existing models.
A1: A generative AI tool designed for legal professionals, featuring capabilities like regulatory chat, document drafting, redaction, and data extraction.
Sarvam Models: A suite of models available via APIs for various tasks including translation (Maurya), speech recognition (Saaras), and text-to-speech (Bulbul).

These models cater to specific applications like bank statement processing (Sesame).

Research and Ecosystem Development

Sarvam AI is committed to fostering a generative AI ecosystem through open-source contributions and large-scale data curation, which will benefit the public good and enhance the capabilities of AI in India.

Sarvam AI aims to democratize AI access across India, ensuring that technology is inclusive and beneficial to individuals from diverse linguistic and socio-economic backgrounds. The company emphasizes building solutions that address the unique challenges posed by India's vast diversity.

Overall, Sarvam AI is positioned as a leader in the Indian generative AI landscape, focusing on making advanced AI technologies accessible and impactful across diverse linguistic and cultural contexts.

Companies