Author(s): Kaustuv Choudhury
Abstract:
As generative artificial intelligence (AI) increasingly shapes the future of global education, top-down ed-tech frameworks face severe implementation hurdles in resource-constrained settings. This paper evaluates the National Digital Education Architecture (NDEAR) in India, analyzing the tension between advanced software blueprints and rural infrastructure realities. While NDEAR's three-pillared digital architecture—Core Registries, Common Interfaces, and Reference Applications—successfully checks private corporate monopolies, its cloud-dependent model collides with severe regional bottlenecks. Drawing on UDISE+ data, we reveal deep disparities in power uptime and network access, alongside an intra-household gender digital divide driven by patriarchal social controls and a linguistic chasm inherent in English-centric language models. To resolve these systemic inequalities, we propose an alternative, democratic engineering blueprint: deploying compressed, quantized Large Language Models running locally on solar-buffered Edge-AI School Server Nodes, integrated with Project Bhashini's open-source multilingual translation pipelines. By shifting digital access from private homes to community-governed village hubs under Gram Panchayat oversight, this framework transforms AI from a centralized corporate luxury into an equitable public good, offering a scalable model for democratic technological sovereignty across the Global South.
PDF URL: View Article in PDF