New Delhi: Global technology consulting and digital solutions provider Tech Mahindra has partnered with NVIDIA to launch a Hindi-first, education-focused Large Language Model (LLM) under its Project Indus initiative.
The new AI model is designed to deliver sovereign, scalable artificial intelligence tailored to authentic Indian linguistic and cultural contexts, with a primary focus on democratizing high-quality education across the country.
Hindi-First AI Model to Support Millions of Students
The newly launched LLM aims to enhance foundational learning in subjects such as physics and other core disciplines. It is built to serve millions of students nationwide, helping them develop deeper conceptual understanding while embracing India’s linguistic and cultural diversity.
The platform is positioned as a comprehensive AI-powered educational resource, supported by multiple partners, to address a wide range of academic use cases.
Powered by NVIDIA’s Advanced AI Infrastructure
The development of the education LLM leverages NVIDIA’s enterprise-grade AI frameworks, including:
- NVIDIA NeMo framework for model training
- NVIDIA NIM microservices for deployment
- Advanced AI infrastructure to ensure production-ready performance and scalability
John Fanelli, Vice President of Enterprise Software at NVIDIA, said the global push for sovereign AI is driving demand for foundation models aligned with local languages and cultural contexts.
“By leveraging NVIDIA AI Enterprise, Tech Mahindra is delivering production-ready performance, reliability and scale to power Project Indus,” he stated.
Addressing Language Data Gaps with Synthetic AI Tokens
To overcome data scarcity in regional languages, Tech Mahindra’s AI team used innovative techniques with NVIDIA NeMo Data Designer to generate over half a billion synthetic tokens. This approach enhanced the model’s contextual understanding in Hindi and other Indian use cases.
The model also supports agentic AI, enabling autonomous agents capable of interacting naturally in Hindi to facilitate AI-driven learning experiences.
Scalable Sovereign AI for India
According to Nikhil Malhotra, Chief Innovation Officer and Global Head of AI & Emerging Technologies at Tech Mahindra, AI is rapidly becoming central to national digital infrastructure and inclusive growth.
He highlighted that many global foundation models are not designed for countries with deep linguistic diversity like India. Project Indus aims to fill this gap by delivering a Hindi-first, sovereign AI model tailored to Indian education and public-sector needs.
Expansion to 8-Billion-Parameter Architecture
Building on its earlier 1.2-billion-parameter foundation model, Tech Mahindra has now scaled Project Indus to an 8-billion-parameter architecture, significantly enhancing model capabilities.
The company is also exploring advancements within the NVIDIA Nemotron ecosystem to:
- Shorten AI training cycles
- Improve instruction quality
- Enhance relevance for Indian education and public-sector applications
Strengthening India’s AI Education Ecosystem
The Tech Mahindra–NVIDIA collaboration marks a significant step in advancing India’s AI-driven education ecosystem. By delivering a sovereign, scalable and culturally relevant AI model, the initiative supports inclusive digital learning and strengthens India’s position in the global AI landscape.
Project Indus is expected to accelerate AI adoption in classrooms while enabling citizen-centric AI services tailored specifically for India.
