Project Repository & Viva Voce Master Kits
Syllabus-aligned Mini & Final Year Major projects complete with synopsis templates, IEEE report structures, PPT slide decks, and external examiner defense questions.
MedVision: AI Multi-Modal Disease Detection & Radiology Assistant
Problem Synopsis: Early diagnosis of pulmonary and oncological conditions is bottlenecked by specialist availability in rural healthcare. MedVision provides an automated triage and heatmap visualization system for healthcare professionals.
Key Architectural Features:
- ✓ Multi-organ pathology detection (Pneumonia, Tuberculosis, Melanoma, Brain Tumors)
- ✓ Grad-CAM visual explainability highlighting activated anatomical regions
- ✓ Automated PDF medical summary generator with confidence intervals
- ✓ HIPAA-compliant encrypted patient record management with role-based access
Top Viva Voce Examiner Questions & Defenses (3 Questions)
Q1: Why did you choose EfficientNet-V2 over standard ResNet-50?
Answer: EfficientNet uses compound scaling (balancing depth, width, and resolution uniformly), achieving 84.3% top-1 accuracy on ImageNet with 4x fewer parameters and significantly faster inference times suitable for clinical environments.
Q2: How does Grad-CAM generate the visual heatmap?
Answer: Grad-CAM computes the gradients of the target class score with respect to the feature map activations of the final convolutional layer. These gradients are pooled to calculate importance weights, creating a coarse localization map highlighting discriminative regions.
Q3: How do you handle severe class imbalance in your medical dataset?
Answer: We implemented Focal Loss with alpha weighting to penalize easily classified negative samples, combined with Albumentations data augmentation and stratified k-fold cross-validation.
📊 View Recommended PPT Slide Structure (8 Slides)
- 1. Introduction & Clinical Motivation
- 2. Literature Review & Existing Solutions
- 3. Proposed Deep Learning Architecture (EfficientNet + GradCAM)
- 4. Dataset Splitting & Preprocessing Pipeline
- 5. Training Metrics (Accuracy, Precision, Recall, ROC-AUC Curves)
- 6. Web App Architecture (FastAPI + React)
- 7. Live Diagnostic Demo & Heatmap Analysis
- 8. Future Scope & Clinical Trial Integration
TrustChain: Blockchain-Based Pharmaceutical Anti-Counterfeiting System
Problem Synopsis: According to WHO, over 10% of medicines in developing nations are counterfeit. TrustChain establishes an immutable custody trail from manufacturer to patient with tamper-proof smart contract auditing.
Key Architectural Features:
- ✓ Smart contract tracking of batch manufacture, distributor handoff, and pharmacy receipt
- ✓ Consumer QR scan verification validating cryptographic hash signatures
- ✓ Temperature sensor IoT telemetry logging over IPFS for cold-chain vaccines
- ✓ Instant counterfeit alert broadcasting to regulatory authorities
Top Viva Voce Examiner Questions & Defenses (2 Questions)
Q1: Why use Polygon/Layer-2 instead of the Ethereum Mainnet?
Answer: Mainnet gas fees ($5-$30 per transaction) make item-level tracking economically unfeasible. Polygon PoS provides sub-cent transaction costs, 2-second block times, and high throughput while retaining EVM compatibility.
Q2: What is the difference between Public, Private, and Consortium blockchains in this context?
Answer: For enterprise pharmaceuticals, a Consortium Blockchain (like Hyperledger Fabric or permissioned Polygon Supernet) allows only verified pharmaceutical stakeholders to write blocks while letting the general public verify.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Problem Statement: Global Counterfeit Drug Crisis
- 2. Blockchain Fundamentals & Smart Contract Architecture
- 3. Sequence Diagram: Manufacturer -> Distributor -> Pharmacy -> Consumer
- 4. Smart Contract Gas Optimization Strategies
- 5. Frontend DApp Integration with Web3 / MetaMask
- 6. Security Analysis (Reentrancy, Front-running mitigation)
- 7. Conclusion & Government Regulatory Feasibility
CodeSync: Real-Time Collaborative Cloud IDE & Sandbox with WebRTC
Problem Synopsis: Remote pair programming and technical interviews require low-latency synchronization and secure execution environments. CodeSync provides collaborative code editing with instant output for 12+ languages.
Key Architectural Features:
- ✓ Real-time multi-user live cursors and color-coded selection highlights
- ✓ Conflict-free replicated data types (Yjs CRDT) ensuring zero merge conflicts
- ✓ Isolated Docker container execution runner with CPU/memory limits and timeout guards
- ✓ Built-in WebRTC peer-to-peer audio/video streaming and interactive whiteboard
Top Viva Voce Examiner Questions & Defenses (2 Questions)
Q1: Why use CRDTs (Conflict-Free Replicated Data Types) instead of Operational Transformation (OT)?
Answer: OT requires a central server to sequence and transform operations, making it complex and prone to latency stalls. CRDTs allow mathematical convergence on all peers without needing a centralized lock-step sequencer, making it ideal for distributed and P2P collaboration.
Q2: How do you achieve sub-50ms latency across connected peers?
Answer: We utilize WebSocket delta compression for text changes and WebRTC P2P mesh data channels for direct audio/video frames, bypassing the central relay server.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Motivation: Collaborative Remote Engineering Tools
- 2. System Architecture: Client -> WebSocket Gateway -> Docker Execution Pool
- 3. Deep Dive into CRDT Algorithm (Yjs state vectors)
- 4. Container Sandboxing & Security Sandbox Hardening
- 5. Performance Benchmarks: Concurrency & Latency
- 6. Live Multi-user Code Execution Demo
- 7. Project Timeline & Deployment Strategy
SmartCampus: IoT Energy Management & Automated Attendance using ESP32 & Facial Recognition
Problem Synopsis: Universities waste up to 35% of power on empty classrooms and lose 10-15 minutes of lecture time on manual attendance. SmartCampus automates appliance power switching and logs real-time attendance to a cloud dashboard.
Key Architectural Features:
- ✓ PIR motion + Ambient Light (LDR) automation for lights and fans
- ✓ Automated attendance logging with 99.2% face recognition accuracy
- ✓ Live power consumption analytics dashboard with predictive anomaly alerts
- ✓ Mobile app for professors to override appliance controls and export CSV reports
Top Viva Voce Examiner Questions & Defenses (1 Questions)
Q1: Why choose MQTT protocol over HTTP REST for IoT telemetry?
Answer: MQTT has a 2-byte header overhead compared to HTTP's 100+ bytes, operates on a lightweight Publish/Subscribe architecture, maintains persistent TCP connections, and supports Quality of Service (QoS 0, 1, 2) levels for unreliable campus Wi-Fi networks.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Campus Energy Waste & Manual Attendance Inefficiencies
- 2. Hardware Components (ESP32, Relays, PIR, Camera Module)
- 3. MQTT Broker & Cloud Telemetry Pipeline
- 4. Facial Recognition Pipeline (Haar Cascades / FaceNet)
- 5. Database Schema & Attendance Ledger
- 6. Hardware Cost Breakdown & ROI Analysis
- 7. Live System Demonstration
ShieldNet: Real-Time Network Packet Sniffer & ML Intrusion Detection System (IDS)
Problem Synopsis: With cyber attacks proliferating, network administrators need lightweight, explainable intrusion detection tools that can inspect promiscuous mode traffic and flag anomalies in real time.
Key Architectural Features:
- ✓ Live promiscuous packet sniffing (TCP, UDP, ICMP, DNS, HTTP)
- ✓ Automated detection of DoS/DDoS, ARP poisoning, and Nmap port scans
- ✓ Trained ML classifier predicting normal vs anomalous flow statistics (NSL-KDD dataset)
- ✓ Visual network protocol distribution and bandwidth waterfall charts
Top Viva Voce Examiner Questions & Defenses (2 Questions)
Q1: How does ARP Spoofing work and how does your project detect it?
Answer: ARP Spoofing involves an attacker broadcasting fake ARP replies claiming their MAC address belongs to the gateway IP. ShieldNet maintains a dynamic ARP cache table and triggers an instant alert when it detects multiple conflicting MACs for a single IP address.
Q2: What is the difference between Signature-based IDS and Anomaly-based IDS?
Answer: Signature-based IDS matches packet patterns against known vulnerability rules (like Snort), which is fast but misses zero-day exploits. Anomaly-based IDS uses Machine Learning to model baseline normal traffic and flags statistical deviations, detecting novel attacks.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Cybersecurity Landscape & Threat Vectors
- 2. Network Protocol Fundamentals (OSI Layer 2/3/4)
- 3. Scapy Packet Capture Architecture
- 4. Feature Engineering from Flow Metrics (Duration, Byte Count, Flags)
- 5. ML Model Evaluation (Precision, False Positive Rate)
- 6. Interactive Threat Visualization Dashboard
- 7. Summary & Future Firewall Integration
FinanceFlow: Personal Expense Tracker & Smart Budget Analyzer
Problem Synopsis: College students struggle to manage monthly pocket money and track discretionary spending. FinanceFlow provides a clean, mobile-first dashboard to log transactions and visualize savings goals.
Key Architectural Features:
- ✓ Secure user authentication with bcrypt password hashing and JWT cookies
- ✓ Interactive Chart.js doughnut and bar charts breakdown of monthly expenses
- ✓ Budget limit warnings with visual progress bars and color-coded alerts
- ✓ Filter transactions by date range, category, and payment mode with CSV export
Top Viva Voce Examiner Questions & Defenses (2 Questions)
Q1: Why do we hash passwords using bcrypt rather than storing them in plain text or MD5?
Answer: Bcrypt uses a slow adaptive hashing algorithm with built-in salting (preventing rainbow table attacks) and configurable work factor (cost) to remain resistant to brute-force GPU attacks.
Q2: What are the core CRUD operations in REST API design?
Answer: Create (POST), Read (GET), Update (PUT/PATCH), and Delete (DELETE), mapping to MongoDB's `create`, `find`, `findByIdAndUpdate`, and `findByIdAndDelete` methods.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Project Objectives & Target User Persona
- 2. MERN Stack Architecture Diagram
- 3. Database Schema Design (Users & Transactions Collections)
- 4. REST API Endpoints Specification
- 5. Key UI Screens (Dashboard, Add Expense, Analytics)
- 6. Security Best Practices (JWT, Sanitization, CORS)
- 7. Q&A and Conclusion
ResQLink: Offline Mesh Network Disaster Relief & SOS Emergency Dispatcher
Problem Synopsis: During natural calamities, telecommunication grids collapse within minutes. ResQLink creates an ad-hoc device-to-device mesh network hopping messages across survivor phones until reaching an emergency relief hub.
Key Architectural Features:
- ✓ Zero-internet P2P SOS beacon broadcasting over Bluetooth & Wi-Fi Direct mesh
- ✓ Offline vector maps caching critical relief shelters and medical stations
- ✓ Triaged survivor intake forms (Medical, Food, Water, Trapped)
- ✓ Relief commander dashboard with automated routing of rescue boats/teams
Top Viva Voce Examiner Questions & Defenses (1 Questions)
Q1: How do messages propagate in a BLE mesh network without internet?
Answer: We utilize Flooding-based Mesh routing. When a phone broadcasts an encrypted SOS packet with a Time-To-Live (TTL) counter, neighboring peer phones receive it, decrement TTL, and rebroadcast it, creating an epidemic dissemination until reaching a coordinator device.
📊 View Recommended PPT Slide Structure (6 Slides)
- 1. The Problem: Telecom Breakdown in Natural Disasters
- 2. Innovation: P2P Bluetooth Mesh Dissemination Protocol
- 3. Mobile App Architecture & Offline-First SQLite Storage
- 4. Triage Algorithm: Priority Scoring of Survivor Needs
- 5. Field Test Results & Battery Consumption Benchmark
- 6. Hackathon Pitch Deck & Impact Potential
CloudForge: Multi-Cloud Automated CI/CD Pipeline & Microservices Cluster on Kubernetes
Problem Synopsis: Demonstrates modern enterprise DevOps engineering practices by containerizing 5 polyglot microservices, managing automated deployments via Helm charts, and maintaining 99.99% uptime with self-healing pods.
Key Architectural Features:
- ✓ Infrastructure as Code provisioning using declarative Terraform scripts
- ✓ Automated blue-green and canary zero-downtime deployment pipelines
- ✓ Prometheus metric scraping with custom Grafana dashboards for latency & CPU alerts
- ✓ Kubernetes Horizontal Pod Autoscaler (HPA) responding to simulated traffic spikes
Top Viva Voce Examiner Questions & Defenses (2 Questions)
Q1: What is the difference between a Pod, a Deployment, and a Service in Kubernetes?
Answer: A Pod is the smallest deployable compute unit (containing 1+ containers). A Deployment provides declarative updates for Pods (managing replicas, scaling, and rolling updates). A Service defines a logical set of Pods and an abstracted network endpoint with load balancing.
Q2: How does Horizontal Pod Autoscaler (HPA) make scaling decisions?
Answer: HPA queries the Metrics Server at regular intervals (default 15s) to calculate current metric utilization against the target threshold: `desiredReplicas = ceil[currentReplicas * (currentMetricValue / targetMetricValue)]`.
📊 View Recommended PPT Slide Structure (7 Slides)
- 1. Evolution from Monolith to Cloud-Native Microservices
- 2. Architectural Blueprint & Network Topology
- 3. Terraform IaC Provisioning Flow
- 4. CI/CD Pipeline Automation (Lint, Test, Build, Deploy)
- 5. Cluster Observability with Prometheus & Grafana
- 6. Disaster Recovery, Chaos Testing & Self-Healing Demo
- 7. Industry Cost Optimization Summary
College Project Submission Kits & Viva Defense Suite
Complete university-ready documentation templates (Synopsis, IEEE 6-Chapter Report, 12-Slide PPT Deck, and External Examiner Viva Defenses) accepted across AKTU, VTU, Anna University, Mumbai University, and Autonomous Colleges.
Project Metadata & Parameters
MedVision: Deep Learning Powered Pulmonary Pathology Detection
1. INTRODUCTION & PROBLEM FORMULATION
Existing diagnostic workflows...
2. OBJECTIVES & PROPOSED METHODOLOGY
3. HARDWARE & SOFTWARE SPECIFICATIONS
Hardware & Software requirements...
4. EXPECTED OUTCOMES & IMPACT
The proposed system will provide transparent, reproducible, and rapid preliminary classification to assist medical practitioners during clinical triaging.
Introduction & Motivation
Defines the problem domain, societal need, project motivation, and boundary scope. Include a 1-page Problem Definition table.
Literature Survey & Related Work
Critical review of 6-10 IEEE/Springer/ACM research papers from the last 5 years comparing approaches, datasets, and limitations.
System Analysis & Architectural Design
Software Requirement Specifications (SRS), Data Flow Diagrams (DFD Level 0, 1, 2), UML Class Diagrams, Sequence Diagrams, and DB Schema.
Implementation & Algorithmic Modules
Detailed breakdown of core algorithms, mathematical equations, API contracts, pseudocode, and key architectural choices.
Testing, Results & Performance Benchmarks
Unit testing (JUnit/PyTest), integration tests, test cases table with Pass/Fail status, confusion matrix, and latency graphs.
Conclusion, Future Enhancements & References
Summary of completed deliverables, limitations of current prototype, 3 concrete future enhancements, and IEEE formatted citations.
Rule #1: The Architecture & Data Flow Diagram (DFD)
Examiners will ask you to draw the DFD (Level 0, 1, 2) or Component Diagram on the whiteboard without looking at slides. Be ready to explain which service calls which database collection.
When asked: 'Walk me through the lifecycle of a single request from clicking Submit on the client to writing to the DB.'
Rule #2: The 'Why Not X?' Trap
Always justify your architectural choices over alternatives. Never say 'Because I only knew React'. Say: 'React's Virtual DOM and component modularity reduced re-render overhead for our live streaming data compared to multi-page static rendering.'
MongoDB vs PostgreSQL: Explain why document schema or ACID relational structure was chosen based on your data relationships.
Rule #3: Code Walkthrough & Live Mutation Test
The examiner might say: 'Change this button color to green', or 'Add a validation so age cannot be negative', or 'Show me where password hashing happens'. Keep your development server running locally with hot reload enabled.
Be ready to open your IDE immediately, navigate to the route handler or component file, and explain the code line-by-line.