From NVIDIA: 2024-10-24 02:15:50

India is a linguistic melting pot with 22 recognized languages and over 1,500 dialects. As digitalization spreads, AI models are being built to allow more Indians to interact with technology in their primary language. NVIDIA has unveiled a small language model for Hindi, with 4 billion parameters, to support localized services across various industries.

Tech Mahindra is using the Nemotron Hindi model to develop Indus 2.0, focused on Hindi and its dialects, for clients in banking, education, and healthcare. The model, showcased at the NVIDIA AI Summit, was fine-tuned with high-quality data to increase accuracy and deliver localized services efficiently.

Companies like Sarvam AI and Gnani.ai are developing AI models for various Indic languages leveraging NVIDIA NeMo. Sarvam AI’s Sarvam 1 supports English and 10 major Indian languages, while Gnani.ai’s speech-to-speech LLM handles real-time voice interactions for banking and financial services. Flipkart, Krutrim, and Zoho Corporation are also building LLMs using advanced NVIDIA resources.

Large enterprises like Infosys, Tata Consultancy Services, and Wipro are utilizing NVIDIA NeMo to develop custom conversational AI solutions for various industries. Combining NeMo with NVIDIA AI Enterprise software enables businesses to enhance customer service interactions and streamline operations with domain-specific language models tailored to their unique needs.



Read more at NVIDIA: Enterprises Build LLMs for Indian Languages With NVIDIA AI