Final Year Project Viva-Voce Master Guide

Model Answer / Defense:Never say 'because it was easy' or 'my guide told me'. Explain based on technical attributes: 'We chose FastAPI and Python for our backend due to its native asynchronous ASGI support, built-in Pydantic data validation, and seamless integration with PyTorch/TensorFlow ML pipelines compared to Django's heavier monolithic footprint.'

Model Answer / Defense:Describe the layers sequentially: 'Our system follows a decoupled 3-tier architecture comprising a client presentation layer (React SPA), a RESTful business logic middleware (Node.js/Express) enforcing JWT authentication, and a persistent transactional data tier (PostgreSQL) coupled with an asynchronous Celery task queue for long-running AI inference.'

Model Answer / Defense:Examiners love this question because it tests whether you actually wrote the code. Detail a specific race condition, CORS header mismatch, or memory leak: 'During concurrent socket connections, multiple users received stale state. We resolved this by introducing Redis pub/sub channels as a message broker between our server clusters.'

Model Answer / Defense:If SQL: 'Our domain required strict ACID transactional guarantees, foreign-key referential integrity between orders, payments, and users, and predictable schema enforcement.' If NoSQL (e.g. MongoDB): 'Our data consisted of semi-structured polymorphic JSON documents (such as dynamic medical records and sensor telemetry) where schema flexibility and high write throughput took precedence.'

Model Answer / Defense:State that you normalized up to 3rd Normal Form (3NF) to eliminate transitive dependencies and insertion/deletion anomalies, with selective denormalization only on high-traffic analytics query paths to avoid expensive 5-table JOIN operations.

Model Answer / Defense:We prevent SQL injection by using parameterized queries through an ORM (Prisma / SQLAlchemy) which separates code from user data. For XSS, our React frontend automatically escapes raw HTML before rendering into the DOM, and our API sets strict Content-Security-Policy (CSP) and HTTP-only cookie headers.

Model Answer / Defense:Passwords are never stored in plaintext. We hash them using the bcrypt algorithm with a salt cost factor of 12. Sensitive environment credentials (database passwords, API secret keys) are injected via isolated server-side environment variables and excluded from git version control.

Model Answer / Defense:Always be honest about limitations—it demonstrates critical engineering maturity: 'Currently, the model requires decent GPU memory to process batch video inputs with sub-second latency, and the payment gateway operates in Sandbox test mode. In future iterations, we plan to containerize inference using TensorRT on edge hardware and add multi-lingual voice navigation.'

Model Answer / Defense:We conducted Unit Testing on individual API endpoints using PyTest/Jest, Integration Testing on authentication and payment workflows, and stress-tested database indexing using simulated concurrent requests via Postman.