Context-aware local querying model using Llama 3 and DeepSeek R1

This research paper presents a novel approach of Retrieval-Augmented Generation (RAG) based model in the field of Generative Artificial Intelligence where we have integrated personalized email data using DeepSeek R1 and Llama 3 in order to run the generative AI model on the local machine using Ollama tool. This approach targets privacy without being totally dependent on the cloud-based servers and gives sub-second query response as compared to traditional solutions. By integrating the personalized email through IMAP protocol, storing the JSON file locally on the device, and further applying RAG based technique using DeepSeek R1 model with 1.5 B parameters. Experimental evaluation explain the decrease in the loss function and quick response times for queries which is less than 1 second, and an accuracy of 89% on testing data. This research emphasized local RAG systems usability in secure, efficient email correspondence management while aiding future development towards privacy-preserving AI communication systems. It focuses on a hybrid approach for accurate query responses of user&s;s emails on their local machines.