Machine Learning Engineer
Current• Spearheading experiments of Reinforcement Learning from Human Feedback (RLHF) to enhance medical logic-building efficiency• Leveraged language models to identify medical documents extraction errors, reaching 90% recall and reducing labelers’ review time by >50%• Fine-tuned LLMs to translate medical guidelines into JSON logics for prior authorization Q&A, boosting labelers’ efficiency by >60%• Designed Reinforcement Learning experiments using medical reviewer feedback to iteratively improve LLM generation• Designed Machine Learning system for medical features labeling using BERT models, resulting in 90% precision and 70% recall• Conducted in-depth data analysis and presented metrics to cross-functional teams, enhancing medical Q&A system performance• Maintained a Python package for medical logic data engineering, ensuring code quality with unit, regression, e2e, and load testing• Implemented and tested internal APIs to abstract complex data engineering processes for medical logic using FastAPI