NEXUS

Inspiration

Exam week. 5 years of question papers. No idea where to start. Every student faces this wall. We thought — what if AI could read all those papers and tell you exactly what to study? That's where NEXUS was born.

What it does

NEXUS is an AI-powered exam intelligence engine. Upload previous year question papers for any subject and NEXUS instantly delivers:

• Repeated questions — exact matches found across multiple years
• Similar question patterns — semantically related questions grouped by theme
• Topic frequency heatmap — which chapters appear most, ranked by weight
• Predicted questions — AI-generated likely questions for your next exam
• Smart study plan — priority order of topics based on historical patterns

How we built it

• Frontend: React + Vite — drag and drop upload, clean results dashboard
• Backend: FastAPI (Python) — file handling and API orchestration
• AI Engine: Gemini 1.5 Flash — reads multiple PDFs natively in a single call, performs semantic analysis, returns structured JSON output
• Infrastructure: Google Cloud (AI LABS BHOPAL2026 project)

Challenges we faced

• Prompt engineering Gemini to return consistent structured JSON across wildly different paper formats
• Handling varied PDF structures across universities and subjects
• Building a results dashboard that makes dense data instantly readable
• Setting up billing and quota management under hackathon time pressure

What we learned

• Gemini's multimodal PDF understanding is a game changer for document-heavy applications
• One well-crafted prompt beats a complex multi-tool pipeline
• Clean output structure matters as much as the AI itself

What's next for NEXUS

• Handwritten paper support via OCR
• Multi-subject cross-analysis
• College-specific question bank repositories
• Mobile app for exam prep on the go

Built With

• fastapi
• gemini-api
• google-ai-studio
• google-cloud
• javascript
• python
• react
• vite