Data Science And Machine Learning Intern
- Implemented and fine-tuned patient risk stratification model for postoperative outcomes. - Improved the performance of classical machine learning algorithms such as logistic regression, random forests and XGBoost on imbalanced data by an average of 10%. - Processed complex and unstructured raw data from multiple sources to form a consolidated and coherent dataset for machine learning. - Implemented and evaluated the effectiveness of various Large Language Models (LLM) such as GPT3.5, Gemini Pro Vision, Mistral, within context-specific Retrieval Augmented Generation (RAG) pipelines. - Improved RAG and LLM architecture for better semantic understanding and answer quality.- Researched and trialled multimodal RAG in a healthcare context, deepening expertise in embedding model architecture and implementations via PyTorch.