IIT Madras partners with Ziroh Labs for new AI research centre, unveils Kompact AI platform

In a move aimed at addressing India’s computing accessibility gaps, the Indian Institute of Technology Madras (IIT Madras) and the IITM Pravartak Technologies Foundation have partnered with California-based startup Ziroh Labs to establish a Centre of AI Research (COAIR). The initiative seeks to develop practical artificial intelligence solutions that operate efficiently on standard CPUs and edge devices, sidestepping the need for costly and hard-to-source GPUs.

IIT Madras and Ziroh Labs have unveiled the first version of Kompact AI, a platform designed to build and run foundational AI models using CPUs. The platform is intended to align with India’s “AI for All” mission, with the goal of making AI tools more accessible and scalable across diverse settings, including rural and under-resourced regions.

Focus on Affordability and Accessibility

Ziroh Labs claims to have optimised 17 AI models—including DeepSeek, Qwen, and Llama—to function effectively on CPU architecture. Benchmark tests, conducted in collaboration with IIT Madras, reportedly assessed both performance and accuracy. Unlike many contemporary AI systems that rely heavily on GPU-based infrastructure, Kompact AI is designed to function offline, making it a potential fit for areas with limited internet connectivity.

The demonstration event was attended by several notable figures from academia and industry, including IIT Madras Director Prof. V. Kamakoti and Turing Award laureate Dr Whitfield Diffie, along with other prominent researchers and industry veterans.

Bridging the AI Divide

Prof. Kamakoti emphasised the importance of focusing on domain-specific models over sprawling, general-purpose systems. "Nature teaches us to acquire knowledge in limited domains effectively," he remarked, suggesting that more targeted approaches are likely to yield sustainable results.

Echoing this sentiment, Dr Madhusudhanan B of IITM Pravartak underscored the ambition to turn research into solutions for communities traditionally left behind by high-cost technology. By leveraging CPU-based AI, the centre aims to develop models that can serve practical, localised needs—ranging from agriculture to education.

As AI development globally continues to favour large-scale, resource-intensive models, the IIT Madras and Ziroh Labs collaboration represents a shift towards more decentralised, energy-efficient alternatives. For India, where infrastructure gaps remain a significant barrier, such initiatives could help democratise access to AI-powered tools and applications.