HOMI - Helpful Online Mentor Interface
Released: Feb 2025
Description
This project is a full-stack application that leverages a fine-tuned Large Language Model (LLM) to provide AI-powered assistance for objective-based learning, specifically for BSc. IT students. The application aims to enhance the learning experience by offering personalized study materials, dynamically generating MCQs based on proficiency levels, resolving user doubts in real-time through a chatbot, and providing tools for assignment tracking, and attendance logging. This system seeks to address the limitations of traditional teaching methods by offering adaptive and interactive educational tools, making learning more accessible, engaging, and effective for IT students.
Key Responsibilities
Assumed the responsibility of "Project Manager" and managed a team of 6 people working on the project
Trained the LLM (based on DeepSeek R1) on cloud and local machines.
Collected the data for training the model from textbooks, using manual scanning, distillation and several scraping techniques.
Developed several backend features, especially those related to interfacing with the actual model.
Assisted in deploying the backend onto HuggingFace Spaces, and uploaded the model onto HuggingFaces.
Assisted in designing several UI components and pages in the application.
Funded the development of the application until a grant was offered to us by our college.
Technology Used
NextJS
TailwindCSS
Typescript
ReactQuery
Django
PostgreSQL
DeepSeek R1
Gemini
Docker
Features
LLM Integration
This system utilizes a customized LLM (based on DeepSeek) for core AI functionalities such as query processing and question generation.
Chatbot Query Resolution
The user can chat with the model to ask questions, generate personalized study material and store and organize them with a robust folder structure system.
Attendance and Assignment Tracking
Users can track and log their attendance and calcuate the number. The user can also log their assignments and use the website as a checklist for these assignments.
Test Generation
The user can assess their knowledge about a certain topic they have studied in the form of graded multiple-choice based tests and as flashcards.
Secure Authentication
The application uses secure authentication methods to log into the system, and a robust database to store user details, questions, tests and more.
Screenshots
Additional Information
This capstone project spanned seven months from ideation to presentation and was presented before my graduation.
Features like a mental health guide, study roadmap generator, and PDF chat export were planned, but they were scrapped due to time constraints.
The project successfully attracted an external investor grant, and these investors plan to further expand its capabilities.
The research done on this project yielded 3 research papers. One was presented at an IEEE conference, one was published in the IJERT journal and one is currently under review.
The project's technical sophistication and innovation were highly praised by an examiner from IITB (Indian Institute of Technology - Bombay) during its final assessment.