Indian GenAI startup Gnani.ai has launched a speech-to-speech large language model (LLM) that boasts of handling over 10 Mn voice interactions daily. The company has stated that the model is being built using NVIDIA’s AI-accelerated computing platform and supports 14 languages.

The LLM was trained on NVIDIA Hopper GPUs using the NVIDIA NeMo platform for developing custom GenAI on over 14 Mn hours of proprietary multilingual conversational data.

“The speech-to-speech LLM is transforming the way enterprises handle their most critical customer interactions. By combining NVIDIA accelerated computing with our conversational AI expertise, we are helping businesses achieve scale and efficiency with a business impact of over $6 billion,” said Gopalan.

The model integrates NVIDIA software, including TensorRT-LLM, Triton, and Riva for real-time performance. Currently, Gnani.ai collaborates with over 150 enterprises across India and the U.S., primarily in the banking, financial services, and insurance sectors.

The launch is in tandem with NVIDIA’s increased focus on India’s AI sector. At the recently held NVIDIA AI Summit in Mumbai, CEO Jensen Huang announced partnerships with over 2,000 Indian startups, including Gnani.ai to develop large language models for Indian languages.