This talk with Noah Rahman, Dustin Hughes, and Kaju Sarkar will focus on the integration of BERT models and advanced audio extraction techniques to for clinical note generation in EMS and explore the challenges of audio data capture in the field. By leveraging domain-specific adaptations we can enhance the accuracy of real-time transcriptions amidst the chaotic environment of ambulances. The presentation will address the unique challenges faced in EMS, such as background noise and speaker differentiation, and demonstrate how AI-powered solutions can streamline documentation, reduce cognitive load on paramedics, and improve patient care through precise and timely medical records.
Kaju Sarkar holds a Bachelors in Mathematics from LSU, an Associate in Computer Science from BRCC and a Bachelors in Film Theory from UNO. He has been programming for over twenty years, in various languages such as C++, JavaScript and Java. He trained at LSU FETI as an EMT, hence he is in a unique position for understanding this problem. Since graduating college, he has been deeply interested in applying Machine Learning, specifically Natural Language Processing, to tasks that require automation but resist a rule based approach. His other area of research is mechanistic interpretability. His ultimate goal for Skribh is to build an open-source foundation model for EMS.
