Riya Das

Traditional academic coaching centers often rely on fragmented systems for attendance tracking, classroom management, support sessions, payments, and communication. Students require real-time academic support, teachers need organized workflows, and administrators need reliable payment tracking and academic records. The platform must support concurrent users while maintaining data integrity, security, and scalability.

Next.js for SSR, SEO, and frontend performance. Node.js & Express.js for scalable event-driven backend architecture. TypeScript for type safety and compile-time validation. PostgreSQL for ACID-compliant relational data management. Prisma ORM for type-safe database operations. Redis for high-speed caching and OTP storage. JWT & bcrypt for secure authentication and RBAC. Google Calendar & Meet APIs for automated scheduling and virtual sessions. Docker Compose for consistent multi-service deployment.

Challenge A: Concurrency Control in 1-to-1 Session Booking. Solution: Implemented atomic database transactions using prisma.$transaction(). The booking process updates an availability slot only when isBooked = false. If another request books the slot first, the transaction affects zero rows and automatically aborts, guaranteeing that a single slot can never be reserved twice. Impact: • Eliminated double-booking scenarios • Guaranteed booking consistency • Improved scheduling reliability under concurrent traffic Challenge B: Redis-Based OTP Caching & Database Load Reduction. Solution: Implemented Redis as an in-memory cache with automatic TTL expiration. OTP generation, validation, and expiration occur entirely inside Redis, achieving O(1) lookup performance. Impact: For approximately 350 OTP requests per day: Without Redis: • 1,050 PostgreSQL operations/day - INSERT - SELECT - DELETE With Redis: • 0 PostgreSQL operations for OTP lifecycle • Reduced OTP-related database operations by nearly 100% • Achieved O(1) in-memory OTP lookup performance Challenge C: Real-Time Classroom Support Queue Management. Solution: Built a concurrent queue management system with active session validation and atomic serial assignment. The system maintains exact chronological ordering while preventing race conditions during simultaneous queue entries. Impact: • Prevented race conditions during queue entry • Preserved accurate student ordering • Improved teacher workflow efficiency Challenge D: Secure Payment Verification & Idempotent Processing. Solution: Implemented server-side payment verification where transaction status is independently verified before updating tuition payment records. Added idempotent processing logic that checks whether a payment has already been marked as PAID before performing any database operation, preventing duplicate processing. Impact: • Eliminated client-side payment manipulation • Prevented duplicate payment processing • Maintained financial consistency. Challenge E: Google Calendar Token Expiration & Meeting Automation. Solution: Designed an automatic token refresh workflow. When API requests fail due to token expiration, the backend securely retrieves the refresh token, generates a new access token, updates stored credentials, and retries the operation transparently. Impact: • Improved scheduling reliability • Reduced failed meeting creation requests • Enabled uninterrupted Google Calendar integration. Challenge F: Containerized Deployment Consistency. Solution: Containerized the complete infrastructure using Docker Compose, including Next.js, Express.js, PostgreSQL, Redis, and Nginx. All services run in isolated containers with consistent configurations across development and production environments. Impact: • Eliminated environment-specific bugs • Simplified deployment workflows • Improved infrastructure portability

Traditional social media feeds prioritize global viral content, drowning out hyper-local notices. Passionate neighborhood experts, local guides, and community educators lack standard micro-transaction systems to monetize premium guides, tutorials, or updates directly with their neighbors.Traditional networks leading to spam. Neighbors require secure, trustworthy content

Next.js was selected for SSR and SEO.Node.js & Express.js for event driven, non-blocking I/O and allow high concurrency on a single thread. TypeScript prevents complie-time discrepancies. PostgreSQL was chosen for ACID-compliance database. Gemini 2.5 flash API selected for generate responses & act as Neighbo-AI.SSL Commerz sandbox supports payment gateway. Google OAuth for user authentication.

Challenge A: Dynamic Algorithmic Feed Weighting & Fast Database Query. Solution: Implemented a Weighted Score Engine for feed ranking and limited feed generation to the latest 100 posts and 50 shared posts before ranking. Impact: • Improved feed generation performance • Reduced database query overhead • Prioritized high-quality and relevant neighborhood content • Delivered faster feed loading for users Challenge B: Concurrency & Callback Safety in SSLCommerz Payment. Solution: Implemented SSLCommerz payment validation using val_id. Wrapped database operations inside prisma.$transaction() to ensure atomic execution. Used status checks (status !== COMPLETED) and upsert operations to prevent duplicate processing. Impact: • Prevented duplicate unlock records • Eliminated callback race conditions • Improved financial data consistency • Protected against fraudulent payment callbacks Challenge C: Handling Multiple Transaction Types. Solution: When a payment is initiated, it generates a unique transaction ID (tran_id) starting with P_ for plans or U_ for post content unlocks. The backend inspects the transaction prefix during callback processing and routes the transaction to the appropriate business workflow. Impact: • Eliminated transaction routing ambiguity • Simplified payment processing logic • Reduced account mapping errors Challenge D: PDF Invoice Generation Compatibility. Solution: Switched from html2canvas to jsPDF. Using jsPDF's native drawing API (roundedRect, text, line, etc.), invoices are generated as vector-based PDFs instead of DOM screenshots. Impact: • Eliminated PDF generation failures caused by modern CSS color functions • Generated sharp, scalable vector PDFs • Reduced PDF size to typically under 50 KB • Improved download and storage efficiency Challenge E: Gemini Chat Initialization Compatibility. Solution: The backend inspects the first item in the chat history array. If the first message has the role model, it is automatically removed before submitting the payload to Gemini. Impact: • Prevented Gemini API initialization errors • Improved AI chat reliability • Ensured consistent chat session creation