In a major step toward making artificial intelligence more accessible, IIT Madras, in collaboration with Ziroh Labs and the IITM Pravartak Technologies Foundation, has launched a Centre of AI Research (COAIR) and unveiled the first version of a new AI platform—Kompact AI.
Designed to run foundational AI models on CPUs instead of GPUs, Kompact AI aims to bridge the computing accessibility gap in India. This platform is particularly beneficial for offline use in rural and low-connectivity areas, aligning with the national “AI for All” mission.
What is Kompact AI?
Kompact AI enables AI model development and execution without requiring expensive GPUs. Ziroh Labs claims to have optimized 17 popular AI models, including DeepSeek, Qwen, and Llama, for CPU environments. In joint benchmarking tests with IIT Madras, the platform demonstrated solid performance and accuracy—paving the way for affordable, scalable AI solutions.
Domain-Focused, Community-Driven
At the launch event, IIT Madras Director Prof. V. Kamakoti highlighted the shift from massive, general-purpose models to domain-specific AI designed for real-world use cases. “Nature teaches us to acquire knowledge in focused domains,” he noted.
Dr. Madhusudhanan B from IITM Pravartak added that the centre aims to convert cutting-edge research into practical tools for communities often left behind by costly tech—like farmers, educators, and local healthcare workers.
A New Direction for AI in India
With the global AI industry leaning toward GPU-heavy, resource-intensive models, the IIT Madras–Ziroh Labs collaboration presents an energy-efficient and inclusive alternative. By focusing on CPU-powered, offline-capable solutions, this initiative could democratize AI for millions, especially in under-resourced regions.
The platform launch was attended by academic leaders and AI pioneers, including Turing Award winner Dr. Whitfield Diffie, showcasing the growing momentum behind India’s push for decentralised, affordable AI innovation.
