56
/
3
Min Read
Tech Mahindra, a global provider of technology consulting and digital solutions, has joined hands with NVIDIA to introduce a new 'Hindi-first Education Large Language Model (LLM)' under the ‘Project Indus’ initiative. This model has been specifically developed keeping in mind India’s education system and linguistic diversity. According to the company, its aim is to provide students with high-quality digital education in their own language, especially Hindi.
This new AI model is focused on the education sector and is designed in such a way that it can help students understand core subjects like Physics, Mathematics and others in simple Hindi language. Its goal is to promote “digitally and linguistically inclusive education” in the country, so that the English language barrier does not become an obstacle in learning.
Project Indus and the Hindi-First AI Initiative
Project Indus is a major AI initiative of Tech Mahindra, developed with the objective of creating sovereign (locally developed) AI models for Indian languages. This model supports Hindi and several of its dialects and is designed for multiple use cases, including education.
This model has been developed using NVIDIA’s AI technology, including the NVIDIA NeMo framework and NIM microservices. These tools have helped in training, scaling and deploying the model, making it suitable for large-scale use.
Data Challenges and Technical Improvements
To overcome the shortage of data for Indian languages, the development team used NVIDIA NeMo Data Designer to create large-scale synthetic training data. According to reports, around hundreds of millions (approximately 500 million) synthetic tokens were generated, which improved the model’s ability to understand language.
This model also supports Agentic AI capabilities, meaning it can create smart AI agents that can answer students’ questions and interact with them in natural Hindi conversation.
Model Scale and Development
Tech Mahindra has scaled this model from an initial 1.2 billion parameter version to approximately an 8 billion parameter architecture, which has significantly improved its capability and performance.
According to NVIDIA, the global demand for “Sovereign AI” is increasing AI systems that are built according to local language and cultural context. This initiative marks a strong step in that direction in India.
Conclusion
Thus, Tech Mahindra and NVIDIA’s Hindi-First Education LLM is a language-focused AI education model for India, designed to make digital learning easier, more accessible and more inclusive. This model is specifically aimed at providing a better learning experience for Indian students in their native language.
