Medical Q&A AI Training and Annotation
I worked on training AI models for Question Answering about Health by doing tasks such as annotating (providing a label to) and assessing medical queries and their answers. I labeled query types as pertinent to the user type (i.e., patient-style), identified the topicality of the query (i.e., symptom, treatment, medication), and evaluated the level of accuracy, clarity, and safety of the AI-produced responses. I worked to eliminate misinformation and to ensure that AI-produced responses adhered to responsible artificial intelligence standards, especially when addressing sensitive health issues. I rewrote or rephrased AI-produced responses to improve their ease of understanding for individuals without medical expertise, while maintaining the factual correctness of the responses. I have managed a large database of medical Q&A pairs and I have implemented a series of strict quality controls to verify guideline compliance, consistency, and extensively document any errors (i.e., inappropriate recommended treatment, omission of disclaimers, unsafe treatment suggestion).