Work, in detail.

A particle swarm optimization approach to mitigating congestion in smart grid AMI (Advanced Metering Infrastructure) networks during post-outage mass reconnection — the surge of devices all trying to reconnect at once after power is restored.

A multi-model medical computer vision pipeline for oral health analysis, combining ResNet18 and DenseNet169 for classification with YOLOv8 + SAHI for detection, and a retrieval-augmented generation layer for clinical reporting. I worked as the RAG Integration & Clinical Reporting Specialist.

• Built the RAG pipeline on the Gemini API for report generation
• Built the canvas rendering engine for annotated findings
• Built the PDF clinical report generator
• Integrated the Brevo email API for report delivery
• Produced an individual report and an A1 poster

Semantic Segmentation and Neural Network Classification of Pro-Myelocytes

• Engineered a Python image processing pipeline to automatically segment pro-myelocytes (PMYs) from a 16,000-image medical dataset
• Applied color space conversions, noise reduction, and advanced thresholding algorithms to generate precise binary masks without requiring manual user intervention
• Interfaced with a neural network SDK in Google Colab to process the segmented outputs for automated, pixel-wise semantic classification
• Quantified pipeline accuracy using mIoU, Dice, and Precision metrics, culminating in a 6-page critically evaluated conference paper

Extended a Java/JavaFX Tetris game with new mechanics (Laser Drill, Time Rewind) and achieved full JUnit test coverage while refactoring the architecture using MVC design patterns.

• Applied Factory, Observer, and Singleton design patterns
• Designed all power-ups, combo systems, UI flow accross six screens, and a persistent leaderboard with JSON-backed score storage

Applied genetic algorithms, particle swarm optimization, and simulated annealing to feature selection for a network intrusion detection system, evaluated on the CICIDS2017 dataset.

• Produced an IEEE-format conference paper with cross-checked results
• Wrote and recorded a video presentation comparing GA, PSO, and SA for IDS

Led the creation of M4TCH, a fast-paced tile-matching desktop game developed in Java using libGDX (LWJGL3) game development framework

• Developed for MacOS and Windows
• Designed all power-ups, combo systems, UI flow accross six screens, and a persistent leaderboard with JSON-back score storage