StudyMate AI — Your Personal AI Learning Companion 🎓🤖

💡 Inspiration

Students today learn from a variety of resources, including textbooks, PDFs, presentation slides, and lecture videos. However, finding relevant information, understanding difficult concepts, identifying weak areas, and knowing what to study next can be challenging.

General-purpose AI assistants can explain concepts, but students also need answers grounded in their actual learning materials, meaningful source references, personalized practice, and a way to measure their progress.

We wanted to explore a better approach: What if one intelligent learning platform could connect study materials, tutoring, assessments, and personalized revision?

That idea became StudyMate AI.

🚀 What It Does

StudyMate AI is an AI-powered learning companion designed to help students understand concepts, practise effectively, and track their learning progress.

📚 1. Multimodal Learning Library

Students can upload educational resources in PDF, PPTX, and TXT formats, along with supported lecture videos.

The platform extracts educational content, organizes it into searchable chunks, and preserves source locations wherever available.

For videos, StudyMate AI processes educational content into timestamped concept segments, allowing students to refer back to relevant moments in a lecture.

🤖 2. AI-Powered Tutoring

StudyMate AI combines information retrieval with Google's Gemini API to support AI-assisted explanations based on available learning materials.

The system retrieves relevant evidence and can attach references to document pages, presentation slides, or video timestamps.

When available evidence is insufficient, the system is designed to identify that limitation rather than inventing source citations.

We are continuing to refine retrieval relevance and answer quality to make tutoring more consistent across different questions and learning contexts.

📝 3. Smart Quiz Generator

Students can generate assessments from available educational materials.

The system supports:

Multiple-choice questions
Conceptual short-answer questions
Numerical questions
Different difficulty levels
Source-linked explanations and feedback

Multiple-choice questions are evaluated deterministically, numerical answers support appropriate tolerance checks, and conceptual answers can be evaluated against a defined rubric using the AI model.

🧠 4. Adaptive Mastery Tracking

StudyMate AI uses a Bayesian Beta-Binomial model to estimate topic-level mastery from recorded assessment attempts.

Rather than treating every student as equally proficient, the system updates its estimates as new answers are evaluated.

It distinguishes insufficient assessment history from demonstrated learning difficulties.

🎯 5. Personalized Revision Recommendations

The platform uses assessment history and estimated topic mastery to identify areas that may need more practice.

Recommendations can target weak topics, revisit misconceptions, and suggest more challenging practice when performance supports progression.

Our broader goal is to make learning increasingly personalized to each student's educational level, selected subjects, and learning needs.

🛠️ How We Built It

We built StudyMate AI using a React, Vite, and TypeScript frontend connected to a Python FastAPI backend.

SQLite stores document metadata, searchable content chunks, quiz sessions, assessment attempts, and mastery information.

The document-processing pipeline extracts content from PDFs, presentations, and text files. Educational videos are processed through Google's Gemini Files API, with timestamped concept segments integrated into the retrieval pipeline.

For document retrieval, we implemented BM25-based lexical ranking to find relevant evidence. The retrieved context is then used by the AI tutoring and quiz-generation workflows.

We also developed structured quiz validation, answer-evaluation logic, a Bayesian mastery engine, and automated evaluation scripts.

🧩 Challenges We Faced

One of our biggest challenges was connecting multiple components into a consistent learning workflow.

Extracting content from different file formats required preserving source locations. Video learning introduced another challenge: answers and generated questions needed to refer to meaningful timestamps rather than simply naming the video.

We also had to consider how to prevent unsupported answers, validate generated questions, evaluate different answer types, and update mastery estimates without inventing student progress.

These challenges taught us that an educational AI system needs more than a conversational interface. Retrieval quality, evidence grounding, assessment correctness, and transparent progress tracking are equally important.

📊 Evaluation

We developed a reproducible assessment evaluation suite and tested several core behaviors.

In our current pilot evaluation:

7 assessment tests passed.
3 out of 3 manually selected gold evidence items were verified.
5 out of 5 tested adversarial, out-of-domain queries were refused in the evaluated grounding workflow.
Deterministic scoring and Bayesian update behavior passed the implemented checks.
The tested set of stored questions had no duplicate prompts.

These results describe a small prototype evaluation, not a guarantee of accuracy across all subjects, learners, or questions. We continue to refine the tutoring experience and expand testing.

💭 What We Learned

Building StudyMate AI helped us understand the importance of combining generative AI with structured data, retrieval systems, assessment logic, and measurable learner progress.

We learned that personalization must be based on meaningful learner information and actual performance, not simply different dashboard designs.

We also learned to distinguish an answer that is supported by a learning resource from one generated using general AI knowledge.

🔮 What's Next

Our next priorities are improving tutoring response quality, deepening school- and college-level personalization, expanding evaluation coverage, and making the platform more reliable for real students.

We want StudyMate AI to become a learning companion that helps students not only find answers, but also understand concepts, recognize knowledge gaps, and make better decisions about what to learn next.

StudyMate AI — Learn Smarter. Understand Deeper. Improve Continuously.