VidyaRANG: Conversational Learning Based Platform powered by Large Language Model

Abstract

Providing authoritative information tailored to a student's specific doubt is a hurdle in this era where search engines return an overwhelming number of article links. Large Language Models such as GPTs fail to provide answers to questions that were derived from sensitive confidential information. This information which is specific to some organisations is not available to LLMs due to privacy constraints. This is where knowledge-augmented retrieval techniques become particularly useful. The proposed platform is designed to cater to the needs of learners from divergent fields. Today, the most common format of learning is video and books, which our proposed platform allows learners to interact and ask questions. This increases learners' focus time exponentially by restricting access to pertinent content and, at the same time allowing personalized access and freedom to gain in-depth knowledge. Instructor's roles and responsibilities are significantly simplified allowing them to train a larger audience. To preserve privacy, instructors can grant course access to specific individuals, enabling personalized conversation on the provided content. This work includes an extensive spectrum of software development and product management skills, which also circumscribe knowledge of cloud computing for running Large Language Models and maintaining the application. For Frontend development, which is responsible for user interaction and user experience, Streamlit and React framework have been utilized. To improve security and privacy, the server is routed to a domain with an SSL certificate, and all the API key/s are stored securely on an AWS EC2 instance, to enhance user experience, web connectivity to an Android Studio-based mobile app has been established, and in-process to publish the app on play store, thus addressing all major software engineering disciplines