As a self-employed Data Scientist in AI & Machine Learning, I develop Python-based solutions for machine learning and data analysis, and build, test, and evaluate predictive models.
I built a multimodal fashion recommendation system using image-text embeddings and user preferences to retrieve and rank clothing recommendations from a dataset of over 250,000 fashion items. Its Transformer-based outfit compatibility model achieved 85.4% test accuracy.
I also developed a multimodal medical AI pipeline combining chest X-ray features from ResNet-50 with clinical text representations from BioBERT. I integrated an AI-assisted inference and medical report workflow using Agno and Streamlit, and implemented Grad-CAM visual explanations.
My work includes analyzing model performance, investigating challenging cases, and documenting technical findings. I use Python, PyTorch, Scikit-learn, Pandas, and NumPy across machine learning and data analysis projects.
