Vedrana I
- Research Program Mentor
PhD candidate at University of California Los Angeles (UCLA)
Expertise
AI applications in medicine, computer science, machine learning, deep learning, computer vision, biomedical data processing and analysis
Bio
Hello! My name is Ved and I am a PhD student in Medical Informatics at UCLA. I focus on applications of machine and deep learning (AI frameworks) to medical questions. Specifically, I explore how we can use these frameworks for medical tasks where we have minimal or no labels for the question we are answering. Currently I work on a variety of medical questions including improving thyroid cancer diagnosis from routine diagnostic imaging as well as health monitoring through wearables (e.g. a smart watch). In my free time I love to bake, explore new coffee shops with a book in hand, and play world building board games. I have mentored undergraduates and early PhD students in the past and enjoy hearing the new ideas and perspectives they bring to solving questions. I look forward to exploring your ideas and scientific questions!Project ideas
Can LLMs interpret time series wearable data?
Wearables like smart watches and rings provide dense data passively collected from our daily life. The resulting time series data is rich with signal that can alert us to early signs of health problems such as heart failure or even if we are coming down with a cold. Making specific models for this data though is difficult and requires a large amount of data and resources. But perhaps we can use LLMs like ChatGPT to interpret this data and understand the trends if we can present them properly. Can we synthesize the dense data into representative features and engineer a well structured prompt that an LLM can understand to predict health events such as stress levels or activity type. If we can, this presents an efficient and straight forward way for every day users of wearables to interpret changes in their physiological health for improved personal health monitoring. This project would be a great exploration of data processing techniques and how to work with the AI models many of us use daily. The results could be presented in a research paper and an app could be created for users of wearables to interpret their own data.