StudyMate — An AI Study Partner Built for a Friend

StudyMate is an AI-powered study partner designed to help a friend manage lecture notes, PDFs, and study material, particularly before exams. Instead of relying on multiple tools, StudyMate consolidates all these resources into one interactive platform. With StudyMate, users can:

- Upload and organize their study materials

- Ask questions about their notes

- Receive answers grounded in their study material

- Generate quizzes for practice

- Create personalized study plans

- Revise important concepts

- Track their learning progress

The primary objective is to transform scattered study material into a personalized, interactive AI study partner. This project is a solution to the common challenge of having abundant study resources but lacking a structured approach to learning, practicing, revising, and preparing.

StudyMate was built using Next.js, React, and TypeScript, emphasizing a modular architecture for AI, document processing, retrieval augmented generation (RAG), database operations, and study features. The AI layer utilizes Google Gemma 2 2B, an open-weight model, which runs locally via Ollama. The workflow involves:

- Accepting study material

- Processing documents

- Chunking and retrieval

- Extracting relevant context

- Generating responses through Gemma 2 2B

This AI architecture forms the foundation for features like AI study chat, RAG-based answers, quiz generation, personalized study planning, revision assistance, and learning progress tracking.

One of the key aspects of StudyMate's development is a provider-based architecture, enabling the application to be decoupled from a single AI model or provider. This design choice allows for easier experimentation with various open-weight models in the future. Additionally, AI coding agents were employed during development to assist with implementation, debugging, refactoring, and organizing the codebase.

The open-weight model, Gemma 2 2B, running through Ollama, serves as the core intelligence behind StudyMate, rather than merely adding AI as an isolated feature. This approach provides several advantages, including the ability to experiment with open-weight AI models, run the model locally, customize prompts and AI behavior, build a RAG pipeline, change or experiment with different models, create a modular AI provider architecture, and maintain greater control over the AI layer's development.

Open innovation holds significance in this context as it grants developers increased control, flexibility, and freedom to experiment with the technology powering their products. In StudyMate's case, this is especially crucial because the application is tailored to students' personal study materials, such as lecture notes, PDFs, and assignments.

By leveraging an open-weight model like Gemma 2 2B, StudyMate can build its AI layer around a model that can be run and experimented with locally, rather than being constrained by a closed API. This flexibility and adaptability are essential in creating a product that caters to the unique needs of individual learners.