StudyBuddy

Inspiration

The inspiration came from the "information overload" students face. Generic AI often gives broad answers, but students need help with their specific textbooks, their specific professor's notes, and their specific exam patterns. We wanted to build a "Digital Twin" of a tutor that knows exactly what is in your backpack.

What it does

Buddy Study is a hyper-personalized AI educational assistant.

Contextual Mastery: It uses Retrieval-Augmented Generation (RAG) to study your uploaded PDFs and notes, ensuring answers are grounded in your specific curriculum.

Command System: Uses a unique "#" command system to trigger specific actions (like #summarize or #quiz-me) instantly.

Gamified Learning: It transforms boring study sessions into "Quests," making academic progress feel like leveling up in a game.

How we built it

Core Engine: Built using the Gemini API for high-reasoning capabilities.

Architecture: Implemented RAG to connect the LLM to private user data safely.

Backend & Deployment: Developed and deployed on Google Cloud Run for scalability and fast performance.

Interface: Designed a streamlined UI focused on "distraction-free" learning.

Challenges we ran into

Data Grounding: Preventing "AI Hallucinations" was tough. We had to fine-tune the RAG pipeline to ensure the AI only used the provided notes when asked, rather than pulling generic (and sometimes incorrect) info from the web.

PDF Parsing: Handling complex layouts, tables, and handwritten notes in PDFs required significant prompt engineering and preprocessing.

Cloud Latency: Optimizing the Google Cloud Run configuration to ensure the response time felt "real-time" during intense study sessions.

Accomplishments that we're proud of

Seamless Integration: Successfully bridging the gap between static documents and an interactive, conversational AI.

The "#" System: Creating a command-based UI that reduces friction and makes the tool feel like a professional developer environment for students.

Successful Deployment: Getting a full-stack AI application live and stable on Google Cloud.

What we learned

Prompt Engineering: We learned that the quality of the AI's output is 90% dependent on how the context is structured.

Cloud Infrastructure: Gained deep experience in managing containerized applications and API quotas.

User-Centric Design: Realized that students don't just want answers; they want a path to understanding.

What's next for StudyBuddy

Multi-modal Support: Adding the ability to summarize YouTube lecture videos and voice notes.

Collaborative Quests: Allowing study groups to join the same "Quest" and compete in real-time quizzes.

Mobile Integration: Developing a dedicated mobile app (building on your interest in app dev) for learning on the go.

Built With

• ai
• aistudio
• chromadb
• docker
• google-cloud
• javascript
• llm
• python
• rag
• streamlit