ARCHIVES

Traditionally, eyecare retailers in tier-2 cities of Nepal (like Janakpur) have had a heavy reliance on brick-and-mortar stores, resulting in fragmented inventory management, limited price transparency and no digital patient records. To counter these systemic issues, we created Chasmawala, a localized e-commerce platform built with Next.js and MongoDB that utilizes a Retrieval-Augmented Generation (RAG) AI system that provides automated, context-aware, and hallucination-resistant customer support in four languages — English, Nepali, Hindi and Maithili. The platform is deployed on the Vercel edge network for low-latency performance optimized for mobile-first users in low-bandwidth rural areas of Nepal. Cross-linguistic semantic embeddings (using paraphrase-multilingual-MiniLM-L12-v2 model) and FAISS-based vector retrieval are used to generate accurate responses specific to eyecare greatly based upon a tailored optical knowledge base. Platform has a Google Lighthouse SEO score of 100 and Best Practices score of 96, demonstrating reliability of architecture. Further features include: digitized eye exams and a specific portal for doctors to access to maintain continuity of eyecare through retail channels. Preliminary evaluation confirms that a localized, AI-driven e-commerce platform can meaningfully reduce information asymmetry, overcome geographical barriers1, and lower digital literacy thresholds for underserved communities — establishing a scalable and replicable model for health-focused digital retail across tier-2 markets in South Asia.