| Year | Project | Built with | Link |
|---|---|---|---|
| 2026 | Personal Portfolio Full-stack portfolio site built with Next.js and Spring Boot, deployed to AWS EC2. Implemented dev/prod environment separation, startup-time data caching, global exception handling, and environment-variable-based CORS configuration. | Next.jsSpring BootTypeScriptAWS EC2Tailwind CSS | |
| 2024 | Glucope - Diabetic Patient Management A web-based patient management application supporting daily workflows for patients and clinicians, deployed to Azure and AWS. | Node.jsExpressJavaScriptHandlebarsMongoDBAzureAWSGitHub Actions | |
| 2024 | Microservices Shopping Backend E-commerce backend built with four .NET Core microservices including Catalog, Basket, Discount, and Ordering, connected via RabbitMQ event-driven architecture with Redis caching. | .NET CorePostgreSQLRedisRabbitMQEF CoreDocker | |
| 2024 | AI Fact Checking System Retrieval-augmented NLP pipeline for climate claim verification, built from scratch under academic constraints. Trained a custom Transformer with Word2Vec embeddings achieving 95% training accuracy, ranked top 10 in class. | PythonPyTorchTF-IDFWord2Vecscikit-learn | |
| 2024 | Deep Image Similarity Matching System Triplet network using fine-tuned ResNet152 for visual similarity retrieval, evaluated with AUC, MAP@K, and t-SNE visualisation. Ranked top 10 on Kaggle leaderboard. | PythonTensorFlowKerasResNet152scikit-learnOpenCV |
2026
Full-stack portfolio site built with Next.js and Spring Boot, deployed to AWS EC2. Implemented dev/prod environment separation, startup-time data caching, global exception handling, and environment-variable-based CORS configuration.
2024
A web-based patient management application supporting daily workflows for patients and clinicians, deployed to Azure and AWS.
2024
E-commerce backend built with four .NET Core microservices including Catalog, Basket, Discount, and Ordering, connected via RabbitMQ event-driven architecture with Redis caching.
2024
Retrieval-augmented NLP pipeline for climate claim verification, built from scratch under academic constraints. Trained a custom Transformer with Word2Vec embeddings achieving 95% training accuracy, ranked top 10 in class.