DSU Research Symposium Features Sureh San
Sureh San, a sophomore currently pursuing a degree in AI presented his research at the DSU Annual Spring Symposium. Sureh’s research was concerned with using Natural Language Processing (a machine learning technique used to analyze and train a corpus of text data) in the context of medical data. The research proposes a novel way to extract valuable medical data using AI to aid medical professionals and augment their efficiency.
Sureh dabbled in both research papers and hands-on experiments to facilitate his research. However, the journey, Sureh says, was not very easy. Inspired by models like Google’s BERT (a language model) and Gemini (Google’s flagship chatbot), Sureh decided to try each one out: both to find out which model works for his specific research goals and to gain a better understanding of the models themselves. In the end, he decided to use a dataset relating to Oncology—a branch of medicine which specializes in diagnosis and treatment of Cancer.
Unfortunately, there were many hurdles in Sureh’s research. The availability of good medical data was scarce, which Sureh attributes to certain existing regulations. In light of the problem with the availability of dataset, Sureh came up with an interesting solution: he decided to synthesize the data. Once the problem with datasets was solved, Sureh tested different pre-trained models to find out which models performed the best. There were also certain restrictions: Sureh had to make sure that the model he chose wouldn’t be too resource intensive. He had to make sure that the model he chose wouldn’t require a huge number of GPUs or memory.
At the end, after going through many different research papers and trying out models, Sureh finally found something that’d work. His research proposes a way to use prior medical data to train a large pre-trained language model. Using that, the research details steps to extract valuable medical data from a user’s medical history. His research makes it easy for medical professionals to accurately get a user’s medical data without missing any important parameters.
Sureh detail’s his experience as being very positive. He thanks his mentor Dr. Mark Spanier for his valuable input and support throughout the research process. Sureh describes his experience with the research presentation as particularly thrilling: there were people with a wide spectrum of knowledge in AI. He says there were people with more experience in AI than him, which made him more passionate about his field. Sureh encourage everyone to participate in research: specifically, he encourages freshmen and sophomores to start researching without fearing anything. With his very positive experience with research and the spring Symposium, Sureh is committed to promoting and enhancing the research quality and quantity from DSU.
