Asian Basic and Applied Research Journal

Maintaining a healthy lifestyle requires proper nutrition awareness, but people frequently find it difficult to decipher food labels, comprehend ingredient compositions, and meaningfully assess nutritional value. The accuracy of traditional nutrition apps is limited when handling a variety of food items, handwritten labels, or blurry images because they rely on static databases and keyword-based matching. In order to overcome these obstacles, we suggest an AI-Powered Nutrition Insight System that combines an image-to-text pipeline, Groq-accelerated LLM inference, and Retrieval-Augmented Generation (RAG). The system uses multimodal models to extract nutritional text from images, embeds that text into vector representations, and uses a RAG module in conjunction with RapidFuzz-based similarity matching to retrieve precise nutrition insights. The system was evaluated using a diverse set of food label images and textual dietary inputs to assess retrieval accuracy, response latency, and contextual reasoning performance. Experimental evaluation demonstrates significant improvements in retrieval precision and response time compared to baseline keyword-based search techniques. The Groq-accelerated inference pipeline further enables low-latency responses, supporting near real-time nutrition analysis and interactive user guidance. Compared to baseline search techniques, experimental evaluation shows big improvements in speed, accuracy, and contextual reasoning. The proposed system therefore provides an efficient, scalable, and context-aware framework for automated nutrition analysis, enabling users to better understand dietary information and make informed food choices.