[{"id":1,"title":"Distributed Event Booking & Payment Microservices System","description":"Architected a fault-tolerant distributed microservices system for university event booking\nwith API Gateway, Event Service, Booking Service, and Payment Service. Implemented\npayment-gated confirmations, retry mechanisms, and service coordination to prevent\noverbooking under concurrent requests. Containerised and stress-tested using Docker\nand PostgreSQL for production-grade backend performance.\n","highlights":["Handled concurrent booking requests with zero overbooking","Payment-gated confirmation with strict correctness guarantees","End-to-end tested using Docker Compose & REST APIs"],"techStack":["Java 17","Spring Boot","Spring Cloud Gateway","PostgreSQL","Docker","Microservices","REST APIs"]},{"id":2,"title":"Banking Microservices with Resilience, Circuit Breaker & Chaos Testing","description":"Designed a resilient banking microservices system with circuit breaker patterns,\nretry logic with exponential backoff, and chaos testing using Chaos Toolkit.\nEvaluated system stability under simulated failures, network delays, and unstable\nbackend behaviour. Reduced API latency from 800ms to 150ms and achieved\nhigh uptime with graceful service fallbacks.\n","highlights":["Reduced API latency from 800ms to 150ms","Circuit Breaker + Retry with Exponential Backoff","Chaos engineering to validate fault tolerance"],"techStack":["Java","Spring Boot","Resilience4j","Docker","Chaos Toolkit","Kubernetes","Distributed Systems"]},{"id":3,"title":"Communication Architecture: Socket vs REST vs gRPC Comparison","description":"Implemented TCP socket communication, RESTful APIs (Spring Boot/Flask), and gRPC\nservices in both Java and Python. Built benchmarking scripts to compare latency,\nscalability, and throughput across protocols. Demonstrated architectural trade-offs\nbetween coupling, speed, and extensibility in distributed environments.\n","highlights":["Benchmarked latency across Socket, REST, and gRPC","Dual-language implementation (Java + Python)","Dockerised all services for consistent testing"],"techStack":["Java","Python","Spring Boot","Flask","gRPC","Docker","TCP Sockets"]},{"id":4,"title":"Distributed Database Replication & Consistency Experiments","description":"Explored replication strategies and consistency models in distributed NoSQL databases.\nConfigured multi-node clusters, experimented with strong vs eventual consistency,\nsimulated node failures, and analysed system behaviour in relation to the CAP Theorem,\navailability, and data integrity across distributed nodes.\n","highlights":["Configured multi-node NoSQL clusters","Tested strong vs eventual consistency under failures","Analysed CAP Theorem trade-offs with real experiments"],"techStack":["MongoDB","Docker","Replication","CAP Theorem","NoSQL","Distributed Systems"]},{"id":5,"title":"Cloud-Native Student Management Microservices System","description":"Developed a containerised CRUD web application using Flask (frontend), FastAPI (backend),\nPostgreSQL (database), and Adminer GUI. Orchestrated using Docker Compose with full\ncreate, read, update, and delete operations and clean microservices architecture\nwith API-based communication between services.\n","highlights":["Full CRUD with Flask + FastAPI microservices","Docker Compose orchestration with 4 services","PostgreSQL with Adminer GUI for data management"],"techStack":["Flask","FastAPI","PostgreSQL","Docker","REST APIs","Microservices"]},{"id":6,"title":"Citrus Disease Detection using Deep Learning (Published – JISEM 2025)","description":"Peer-reviewed research project presenting a CNN-based deep learning system for\nautomated citrus fruit and leaf disease detection. Trained on preprocessed datasets\nwith data augmentation (rotation, flipping, zooming). Achieved 95% accuracy for fruits\nand 99% for leaves. Published in JISEM 2025.\n","highlights":["95-99% model accuracy on disease classification","Published research in JISEM 2025","Real-world AI solution for agricultural diagnostics"],"techStack":["Python","TensorFlow","Keras","OpenCV","CNN","Deep Learning","Image Processing"]},{"id":7,"title":"AI vs Human Text Classification (NLP)","description":"Developed an NLP system to classify AI-generated vs human-written academic abstracts\nusing TF-IDF feature vectorisation. Evaluated Naive Bayes, Logistic Regression, and\nk-NN models with robust validation. Used to estimate AI content proportions in\nunseen test datasets of 1000+ abstracts.\n","highlights":["Multi-model comparison with rigorous evaluation","TF-IDF + Naive Bayes, Logistic Regression, k-NN","Estimated AI content proportions in unseen data"],"techStack":["Python","Scikit-learn","TF-IDF","NLP","Machine Learning","Text Classification"]},{"id":8,"title":"Embedding-Based NLP Classification (BERT, Word2Vec, Doc2Vec)","description":"Extended traditional NLP with embedding-based text classification using semantic\nvector representations. Implemented Word2Vec, Doc2Vec, and BERT models with\nensemble classifiers to improve accuracy and semantic understanding of\nAI-generated vs human text.\n","highlights":["BERT, Word2Vec, Doc2Vec embeddings compared","Ensemble classifiers for improved accuracy","Semantic understanding beyond bag-of-words"],"techStack":["Python","BERT","Word2Vec","Doc2Vec","NLP","Deep Learning","Ensemble Methods"]},{"id":9,"title":"Bitcoin Price Trend Prediction using ML","description":"Built predictive models for BTC price trends using 700,000+ rows of 1-minute\ncandlestick data. Engineered technical indicators (RSI, MACD, moving averages,\nvolatility, momentum). Trained Decision Tree, Random Forest, and XGBoost models\nwith rigorous evaluation across train/validation/test splits.\n","highlights":["Trained on 700K+ financial data points","Engineered RSI, MACD, and momentum features","Precision, Recall & F1-score evaluation"],"techStack":["Python","XGBoost","Random Forest","Pandas","Time Series","Data Mining"]},{"id":10,"title":"Advanced BTC Prediction with Deep Learning","description":"Extended baseline BTC prediction with advanced ML and deep learning models\nincluding a custom improved architecture. Implemented strict dataset splits,\nhyperparameter tuning, and comprehensive model comparison using accuracy,\nprecision, recall, and F1-score across all datasets.\n","highlights":["Custom deep learning architecture design","Hyperparameter tuning with validation monitoring","Comprehensive multi-model comparison"],"techStack":["Python","TensorFlow","Deep Learning","XGBoost","CNN/LSTM","Time Series"]},{"id":11,"title":"AI-Generated Game Code Evaluation with JUnit Testing","description":"Designed a comprehensive JUnit test suite (30+ test cases) to evaluate correctness,\nedge cases, exception handling, and logical flows in AI-generated Java game code.\nConducted qualitative code review identifying design inefficiencies, poor practices,\nand improvement opportunities in OOP implementations.\n","highlights":["30+ JUnit test cases for thorough coverage","Edge case, exception, and boundary testing","Qualitative code review and OOP analysis"],"techStack":["Java","JUnit","Unit Testing","OOP","Code Review","Software Engineering"]}]