Know India
AI-powered tourism platform for exploring India's states and territories.
Overview
Problem
India has 28 states and 8 union territories, each with unique culture, cuisine, and attractions. Travelers struggle to discover authentic experiences and plan comprehensive itineraries without fragmented research across dozens of sources.
Solution
Built a full-stack tourism platform that combines structured geographic data with AI-generated personalized itineraries. Users can explore regions interactively, get tailored travel plans via a RAG pipeline, and export their itineraries as shareable PDFs.
Tech Stack
Architecture
Features
- Interactive map-based exploration of all Indian states and territories
- AI-powered itinerary generation using Mistral-7B and FAISS retrieval
- Category-based discovery: food, culture, adventure, heritage
- Save favorites and download itineraries as PDF
- Admin dashboard for managing tourism data
- Responsive design for mobile and desktop
Technical Decisions
Chose Mistral-7B over cloud APIs for cost efficiency and offline capability during development
Used FAISS for vector similarity search to match user preferences with relevant destinations
MySQL for structured relational data (states, districts, attractions) with normalized schema
React with Tailwind CSS for rapid UI iteration and consistent styling
Challenges
- Curating accurate, up-to-date tourism data for all 36 regions with consistent quality
- Balancing LLM response quality with latency — implemented streaming responses
- Designing a RAG pipeline that produces contextually relevant itineraries, not generic content
Outcomes
- Published as an npm package (@aryanjsx/knowindia) with 245+ weekly downloads
- Interactive India map component extracted as a separate reusable package
- Comprehensive coverage of all Indian states and union territories