The IndiaAI–NCG CATCH Grant for Cancer 2026 compendium documents ten winning AI solutions selected through a competitive grant program jointly run by IndiaAI (MeitY) and the National Cancer Grid's Koita Centre for Digital Oncology. It profiles AI tools spanning cancer screening, diagnostics, clinical decision support, patient engagement, and operational efficiency, ranging from pilot-ready to commercially deployed systems. Each solution is evaluated against criteria including clinical relevance, technical maturity, data governance, DPDP compliance, and deployment feasibility within Indian hospital settings. The compendium also covers the selection methodology, jury composition, responsible AI principles, and key learnings from the challenge, serving as a reference for hospitals, policymakers, and innovators seeking to adopt or scale AI-enabled cancer care across India.

Neural Architecture Search has emerged as a transformative paradigm for automating the design of deep neural networks, reducing the reliance on human expertise and enabling the discovery of architectures that often surpass manually designed counterparts. This paper presents a comprehensive survey of NAS methodologies, tracing the evolution from early reinforcement learning-based and evolutionary approaches to modern differentiable and one-shot methods. The paper systematically categorizes NAS techniques along three fundamental dimensions: search space design, search strategy, and performance estimation. A unified mathematical framework is introduced that encompasses gradient-based, reinforcement learning-based, and evolutionary NAS under a common optimization objective.

Why owning both the silicon and the software is the only durable moat in the age of AI, and why the margin arithmetic of fragmented supply chains is finally breaking down. By Dhritiman Mallick · Founder & CEO, Vyuhaa Med Data

AI agents are moving from answering questions to taking real actions inside enterprise systems. This article explains how businesses can design safer AI agents with the right permissions, monitoring, policy checks, and human oversight, so agents can act autonomously while keeping critical decisions under control.

Most teams choose a model using two numbers that were measured separately and never reconciled: a quality score and a tokens-per-second figure. This paper joins them. We measured eight vision-language models twice — ranking accuracy against ground truth, and prefill and decode throughput on real queries — on a single 24GB RTX 3090, one model at a time. The two measurements rank the fleet in near-opposite orders, and most of that turns out to be parameter count rather than a capability trade-off. What survives is more useful: decode throughput measures your serving configuration rather than your model, short structured outputs are almost entirely prefill-bound, and per-request speed can invert against per-job cost by a factor of two and a half. 28 pages, full method, four instrumentation defects we caught before trusting any number, and an extended set of caveats.