The Dashboard

🇮🇳 Prasikshan

Tagline: "Master SSB. Serve the Nation."

🚀 Our Journey Building Prasikshan

💡 What Inspired Us

As a proud Indian, we have always been inspired by the men and women who wear the uniform of our Armed Forces. During college, we saw many of our friends and juniors preparing for the SSB (Services Selection Board) interview — one of the toughest and most holistic selection processes in the world.

Despite having talent and dedication, most aspirants struggled due to:

Lack of quality resources
Expensive coaching
No proper feedback system for psychological tests like TAT, WAT, SRT, and PPDT

This personal observation became the spark.

We wanted to build something that could democratize SSB preparation — making world-class training accessible to every aspirant across India, especially those from small towns and rural backgrounds.

🛠️ How We Built the Project

Prasikshan was built from the ground up as a full-stack, production-ready platform in roughly 6 weeks of intense development.

⚙️ Tech Stack Highlights

Frontend: Next.js 16 (App Router) + TypeScript + Tailwind CSS + Recharts
Backend: Node.js + Express-style API routes with secure JWT dual-token authentication
AI Microservice: Python + FastAPI + Azure OpenAI (Grok-3)
Database: MongoDB Atlas with optimized sharded collections
Infrastructure: Docker, Nginx, Cloudinary CDN, Redis rate limiting

🧠 Architecture Highlights

Implemented HTTP-only cookies with dual JWT authentication
Built complete SSB test suite:
- OIR → PPDT → TAT → WAT → SRT → Lecturette → PI
Designed a dedicated AI microservice for:
- Prompt engineering
- Context-aware evaluation
- High performance isolation from main backend

⚔️ Challenges & Solutions

🔹 Complex AI Evaluation Logic

Problem: Azure OpenAI calls caused header size errors (431) and inconsistent responses
Solution:
- Migrated to a structured FastAPI microservice
- Used rigid prompts and Pydantic-based response parsing

🔹 Performance & Database Bloat

Problem: Storing all results inside the User document slowed queries
Solution:
- Created a separate UserResults collection
- Implemented optimized schema and migration scripts

🔹 Real-time Progress Tracking

Problem: Maintaining test flow and user progress across devices
Solution:
- Built a custom progress engine
- Dynamically serves the next logical test

🔹 Time Management

Balancing college academics with 14–16 hour coding days was extremely demanding.

🔹 Security

Implemented:

Rate limiting
Input sanitization
Secure cookies
Production-level backend safety practices

📚 What I Learned

This project taught me far more than just coding:

Importance of clean architecture and separation of concerns
Designing user-centric systems
Deep understanding of prompt engineering
Building AI systems that simulate real-world experts
Practical DevOps skills:
- Docker
- Environment management
- Scalable architecture design