StudyBuddy

Teacher uploads data like reference material/ quizzes/ code snippets/etc
Login page for professor/ student
One studyBuddy for one course
A personalized touch to conversations
Empathy is a major factor while teaching, keeping in mind the mood of the student
Personalized touch to conversations

Inspiration

As children, many of us experienced moments when teachers or parents lacked patience while helping us understand complex topics, such as mathematics. Understanding these subjects requires different amounts of time for different people, and humans are not infinitely patient. This inspired us to create an empathetic voice agent study partner, StudyBuddy, that ensures students comprehend all class topics, particularly the more challenging ones that are hard to grasp during a single class session.

What it does

StudyBuddy is an AI chatbot controlled and governed by teachers. Teachers provide the voice agent with course materials, quizzes, and other relevant context, making the AI aware of all the nuances of the course. Students can then ask various questions and doubts, which the AI addresses empathetically, considering the student's mood through voice and video input.

How we built it

We developed this application using the OpenAI API, LangChain, and HumeAI for emotion detection and the empathetic voice agent. The frontend was built with React JS. Additionally, we trained a custom LLM on an Intel GPU (phi3 model) and created a Retrieval-Augmented Generation (RAG) agent that can accept PDF/URL inputs to better understand the coursework and the students' needs.

Challenges we ran into

One major challenge was integrating our custom LLM with HumeAI for output.

Accomplishments that we're proud of

We are proud to have created a functional RAG that answers specific course-related queries and developed a user-friendly frontend.

What we learned

We learned a great deal about building RAG agents, model fine-tuning, and numerous other AI concepts and libraries.

What's next for StudyBuddy

Next, we plan to implement the system with our own local LLM, which will better assist with specific materials such as homework and assignments. We will also write code to ensure that answers are not directly revealed; instead, the AI will guide students towards finding the answers themselves.