At an annual event for Software-as-a-Service (SaaS) startups, SaaSBoomi, Vivek Raghavan, the co-founder of Sarvam AI, a homegrown generative artificial intelligence (AI) startup backed by Lightspeed, revealed plans to unveil its inaugural commercial 'voice-to-voice endpoint' within the next six to twelve months. Raghavan highlighted the multilingual capability of this technology, asserting that it would support at least 10 Indian languages and provide various user experiences. Stressing the importance of voice-based, action-oriented language models (LLMs) that cater to colloquial languages, Raghavan emphasized the need for accessibility to a wide audience through voice-driven interfaces.

Sarvam AI's commitment to advancing Indian language AI was underscored by the release of its first open-source Hindi language model, OpenHathi-Hi-0.1, in December of the previous year. This initiative is part of a broader effort by the company to foster innovation in the Indian language AI ecosystem through the provision of open models and datasets. Raghavan also highlighted the superior performance of OpenHathi in English to Hindi translation compared to existing models like GPT-4 and GPT-3, illustrating the potential for customized models to significantly enhance performance in Indic languages.

Despite the progress made, Raghavan acknowledged the challenges facing the development of Indian language AI, including the underrepresentation of Indic languages in training data and the poor quality of available data. He also identified tokenization and evaluation as additional hurdles to overcome. Sarvam AI's ambitious vision and recent funding success, having raised $41 million in its Series A round led by Lightspeed Ventures, position the company as a key player in driving innovation and accessibility in Indian language AI. Founded in July 2023 by Vivek Raghavan and Pratyush Kumar, both of whom have extensive experience in the field, Sarvam AI aims to revolutionize the landscape of AI technology in India.