“The goal is not to build machines that think like humans, but to build machines that help humans think better.” - Demis Hassabis, CEO, DeepMind
Advancing India’s reasoning models
Fractal Analytics, recently selected under the Central Government’s IndiaAI Mission, is focusing on developing AI systems that move beyond knowledge retrieval to reasoning and analytical capabilities. Among its models, Fractal is building an image-to-image diffusion model called Kalaiodo.ai and another reasoning-focused model, Fathom-R1-148B, aimed at strengthening STEM and mathematical reasoning. The company’s approach differs from the first round of IndiaAI proposals, choosing a more research-oriented and science-led direction that supports foundational model innovation.
Problem-solving and scientific impact
Amonkar highlighted that the path to Artificial General Intelligence (AGI) may not rely solely on language. Instead, Fractal’s proposal emphasizes math, STEM reasoning, and medical data, where strong problem-solving capabilities could unlock broader AI understanding. The models aim to enhance cognitive flexibility and analytical reasoning in scientific fields.
Capturing local nuances and datasets
Fractal’s models also seek to capture India’s local nuances, especially in healthcare, where data diversity and treatment guidelines differ from Western standards. For instance, AI models trained on Indian Council of Medical Research (ICMR) datasets for Covid-19 will better reflect India-specific health realities and reasoning patterns.
Building indigenous AI sovereignty
Fractal is developing these models indigenously, ensuring that ownership, training data, and geographic origin remain within India’s AI ecosystem. The company sources data from multiple channels, including synthetic datasets, to build high-quality AI foundations. Amonkar also acknowledged government support for access to GPUs, a critical step toward affordable AI innovation.
Summary
Fractal Analytics’ IndiaAI-backed project aims to advance India’s position in reasoning-based AI, with a focus on math, STEM, and healthcare reasoning. Through indigenous model development, localized datasets, and scientific problem-solving, Fractal is steering India toward greater AI sovereignty and foundational innovation.
Food for thought
Can India’s focus on reasoning-based AI models help it become a global leader in scientific and medical innovation?
AI concept to learn: Reasoning Models
Reasoning models are advanced AI systems designed to understand, infer, and apply logic beyond memorized information. Unlike traditional knowledge models, they can solve problems, analyze contexts, and generate new conclusions, mirroring the way humans reason through complex tasks.
[The Billion Hopes Research Team shares the latest AI updates for learning and awareness. This is not a professional, financial, personal or medical advice. Please consult domain experts before making decisions. Feedback welcome!]
COMMENTS