Tech Mahindra has announced the launch of an education-focused large language model (LLM) under Project Indus, developed in collaboration with NVIDIA. The Hindi-first model aims to support sovereign AI development while expanding access to AI-driven learning experiences tailored to India’s linguistic and cultural context.
The new model is positioned as an education-centric platform designed to democratise access to foundational learning across subjects such as physics and other academic disciplines. According to the company, the LLM is intended to support scalable educational use cases while addressing the diversity of Indian languages and learning environments.
Built using NVIDIA AI infrastructure, the model leverages the NVIDIA NeMo framework and NVIDIA NIM microservices to enable performance, scalability and deployment readiness. To address language data limitations, Tech Mahindra said it generated synthetic training data using NVIDIA NeMo Data Designer, producing hundreds of millions of tokens to strengthen the model’s linguistic capabilities. The platform also introduces agentic AI functionality, enabling the development of autonomous learning assistants capable of interacting in natural Hindi.
The latest version represents a significant scale-up from the earlier Project Indus model, expanding from approximately 1.2 billion parameters to an 8-billion-parameter architecture. The updated model has been designed specifically for education and public-sector use cases aligned with India’s digital learning priorities.
Nikhil Malhotra, Chief Innovation Officer and Global Head of AI and Emerging Technologies at Tech Mahindra, said locally grounded language models are critical to enabling inclusive AI adoption in markets with diverse linguistic and cultural requirements. He added that Project Indus aims to bridge the gap by delivering a sovereign AI model tailored to India’s education ecosystem.
John Fanelli, Vice President, Enterprise Software at NVIDIA, noted that demand for sovereign AI models is increasing globally, with organisations seeking AI systems aligned to local language and regulatory contexts.
Tech Mahindra said future iterations of Project Indus may incorporate capabilities from NVIDIA’s open model ecosystem to accelerate training cycles and improve domain specialisation. The initiative forms part of the company’s broader efforts to develop locally relevant AI solutions supporting India’s digital transformation agenda.