Projects
AI-Based Skin Cancer Detection
Finetuned and evaluated ten CNN variants for systematic comparison of widely used CNN architectures for skin cancer detection.
Synthesized these findings into a qualitative and quantitative analysis for a research paper published in AJSET.
MelanoDetect AI
Led the development of AI-powered handheld melanoma detection device with a mobile app. The project secured HKD 100,000 in funding through the Hong Kong Science and Technology Park incubation program.
Privacy Preserving Sentiment Anlaysis based on various Capitals
Developed FedBiLSTM-Net by integrating LSTM with federated learning using the FedAvg algorithm for privacy-preserving sentiment analysis. Trained the model across 20 non-IID clients over 9 rounds, achieving a centralized evaluation accuracy of 80.17%. Co-authored the resulting research paper, published in the International Journal of Social Sciences Bulletin
LLM Reasoning
Developed a two stage pipeline using the Indiana University- Chest X-Rays dataset for automated radiology report generation by implementing Qwen-VL and MedGemma-27B during my internship at Medical Imaging and Diagnostics Lab, National Center of Artificial Intelligence (NCAI) .
