During his recent trip to India, Microsoft CEO Satya Nadella announced that the technology giant is partnering with Bengaluru-based Indian startup Sarvam AI to integrate its Indic voice large language model (LLM) into the Azure platform.
The collaboration will allow Sarvam AI to leverage the Azure AI Infrastructure for constructing and deploying its voice LLM stack in Indian languages. Additionally, the partnership seeks to equip developers to create generative AI applications on a large scale and facilitate enterprises in the swift and responsible adoption of generative AI.
Pratyush Kumar, Co-founder of Sarvam AI, said, “We are very excited to collaborate with Microsoft to make advanced AI technology accessible and impactful for everyone in India. This partnership embodies our vision of ‘Sarvam’ – meaning ‘all’ – by enhancing AI’s reach across various Indian languages and sectors.”
The collaboration extends beyond technology development, with both companies working together to aid enterprises in adopting generative AI rapidly and responsibly. Sarvam AI will leverage Microsoft's cloud and AI infrastructure, including Azure OpenAI Service and Azure Machine Learning, for efficient training, hosting, and scaling of LLMs.
“We are committed to enabling AI for everyone, empowering India’s transformation into an AI-first nation,” said Puneet Chandok, President of Microsoft India and South Asia.
“Through our collaboration with Sarvam AI, we are not just supporting homegrown innovation – we are fostering a future where every individual, regardless of their language or background, can benefit from the power of voice-driven AI solutions. Together, we are taking a significant step toward enabling India’s people, communities, and organizations to achieve more,” Chandok added.
Sarvam AI, based in Bengaluru, specializes in developing generative AI models tailored for Indic languages and contextual scenarios. Around two months ago, the company launched Hindi LLM OpenHathi-Hi-v0.1. Built on Meta’s Llama2-7B architecture, Sarvam AI claims that it exhibits performance equivalent to GPT-3.5 for Indic languages. The model extends Llama2-7B’s tokenizer by 48,000 tokens and undergoes a two-phase training process.
The company’s goal is to enhance the accuracy and cost-effectiveness of developing and deploying generative AI applications in India. The Indic voice LLM, initially available in Hindi, aims to provide a natural voice-based interface and will expand to cover more Indian languages, supporting colloquial language use.
The Voice interfaces are considered one of the most natural platforms for generative AI applications in Indian languages, applicable in education, finance, healthcare, and customer service sectors.
Microsoft's move to make Sarvam AI's Indic voice LLM available on Azure paves the way for Indian developers to build real-time, voice-driven generative AI applications on a large scale.