Kaththi Sandai

• 

  Main UI

Inspiration What do you remember about KATHTHI SANDAI? In Tamil culture, it has a dual meaning: a legendary sword fight or a chaotic shouting match. For most high schoolers, preparing for JEE, NEET, or Board exams feels like the latter—a stressful shouting match between their brain and their textbooks. I wanted to change the narrative. I wanted to turn the "shouting match" into a literal Sword Fight where students can fight for their goals instead of just suffering through them. What it does Kaththi Sandai is a gamified learning platform that turns exam prep into a battlefield. The Battlefield: The entire UI is designed as an arena. You aren't a "student"; you are a Warrior. Engagement: Instead of boring MCQs, you engage in duels. Progression: Features like Warrior Profiles and a live Leaderboard turn solitary study sessions into a competitive, social experience. Targeted Prep: Specifically built to handle the rigors of JEE, NEET, and high school Board exams. How we built it The project was a high-speed, 7-day sprint: The Brain: We used Gemini 2.5 Flash for its lightning-fast inference and intelligent question generation. The Logic: Integrated with LangChain to orchestrate AI agents and manage the flow of data. The Structure: Used Structured Prompting to ensure the AI strictly outputs JSON, allowing for seamless integration between the AI and the backend. The Team: I (J. Nishanth Raj / BAVA 🗿) handled the Backend and AI Architecture, while Gemini acted as the Frontend Developer. Deployment: Hosted on PythonAnywhere. Challenges we ran into The biggest hurdle was moving from "AI that talks" to "AI that builds." Getting a LLM to consistently output perfectly structured JSON for complex physics and math problems without "hallucinating" took a lot of prompt refinement. Also, building a fully functional web app with a backend and AI integration in just one week meant very little sleep and a lot of debugging. Accomplishments that we're proud of Speed-to-Market: Going from an idea to a live, hosted web app in 7 days. Agentic Workflow: Successfully using LangChain to make the AI act as a reliable question generator rather than just a chatbot. User Experience: Creating a theme that actually makes you want to solve "one more problem" to climb the leaderboard. What we learned I learned that the "Human + AI" team is incredibly powerful. By treating Gemini as a teammate rather than just a tool, I was able to scale my output and focus on the complex AI architecture. I also realized that Gamification isn't just a buzzword; when you change a student's identity from "stressed learner" to "warrior," their productivity sky-rockets.

Built With

• flask
• langchain
• python
• pythonanywhere