Building AI Models in Healthcare Using Semi-Synthetic Data

Learn how to build AI models for healthcare applications using semi-synthetic data in this 40-minute conference talk. Explore the challenges of collecting training data in regulated fields like healthcare and discover how Fight Health Insurance leveraged de-identified public and proprietary information to create semi-synthetic training sets for fine-tuning machine learning models. Examine practical approaches to incorporating the latest reasoning techniques in fine-tuning while developing cost-effective models that can run on single GPU inference rather than requiring expensive A100 clusters. Gain insights into real-world implementation through discussion of the Fight Paperwork system and access to the accompanying public GitHub repository code for hands-on exploration and inspiration.