Askademia

Inspiration

Students waste time hunting through long syllabi for simple answers—Askademia brings a TA’s knowledge to your fingertips.

What it does

Uploads course PDFs, chunks and embeds them in MongoDB Atlas Vector Search, then lets students chat with an LLM-augmented interface that returns precise answers.

How we built it

• Backend: FastAPI + PyMongo vector search + Google Gemini embeddings & generation
• Ingestion: pymupdf + sliding-window chunker + batch uploads
• Frontend: Create React App with a themed chat UI and fetch calls to /query

Challenges we ran into

• Tuning chunk size vs. token limits
• Provisioning and querying Atlas vector indexes
• Configuring CORS and environment variables

Accomplishments that we're proud of

• Fully automated RAG pipeline from PDF → chat answer
• Responsive, dark-mode React UI
• Modular codebase that swaps models or vector stores easily

What we learned

• Real-world MongoDB Atlas Vector Search usage
• Prompt design for grounded LLM responses
• Seamless local/cloud env management with python-dotenv

What's next for Askademia

• Production Docker/Kubernetes deployment
• Multi-course support and query analytics
• OCR ingestion for images/slides and real-time collaboration features