Implementing LLM for E-Learning: A Real Case of AI Integration in a Web Application

AIcademy is an innovative web application in the field of education. Leveraging advanced natural language processing techniques, it enables the creation of virtual avatars that closely resemble human figures, capable of understanding, engaging in conversations, and answering questions. In this project, my team and I developed an e-learning platform where users can create virtual avatars specializing in particular topics or fields and offer them for sale. These avatars are trained on documents provided by the creator, such as PDFs, images, or text files. For example, students interested in learning how to code in Java can purchase an avatar and follow its lessons. Each lesson is interactive, allowing students to ask questions, clarify doubts, and adjust the avatar's teaching style based on their proficiency level. Additionally, students can be tested throughout the learning process. We used a vector database to store word embeddings, an embedding model to process uploaded documents and student queries, OpenAI's large language model to handle requests, and the ElevenLabs API to convert text into speech. This type of learning offers several benefits, such as personalized learning experiences, but it also comes with challenges, such as bias.