This project introduces the SwinColorFusionNet model, which leverages a Swin Transformer combined with color histogram features to enhance the accuracy of Mayo Endoscopic Score classification for Ulcerative Colitis. The proposed model significantly outperforms both a naive baseline and a ResNet-101 based model, achieving a QWK score of 0.8552 and an accuracy of 77.4%.

This project involved developing a custom sentiment classification model which uses a modified GPT architecture. By focusing on hyperparameter tuning and data preprocessing, the model achieved a 77% accuracy, significantly improving the identification of positive and neutral sentiments.

Deploying Transformer-based models for text classification and named-entity recognition (NER), enabling precise document categorization and entity identification.

The project's objective is to develop an AI model that enhances the accuracy of diagnosing common thoracic diseases in children from chest X-rays. By incorporating advanced data augmentation techniques alongside patient age and sex, addressing the challenges of image variability, and improving disease differentiation.

This project investigates the influence of game phase dynamics on match outcomes in League of Legends, utilizing data analysis to understand how key events and strategies impact victory. It delves into the strategic aspects of MOBA gaming, focusing on objective control, eliminations, and economic advantages, aiming to uncover patterns that lead to successful outcomes.

Implementing AI tools for business efficiency, predictive analytics, and data visualization.

Specializing in deep learning, machine learning, and data engineering.

Utilized programming and data science for financial process improvements.

Focused on financial planning and analysis.

Conducted financial performance analyses.