My research centers on developing and evaluating foundational natural language processing (NLP) methods, particularly large language models (LLMs), to convert complex data, such as electronic health records (EHRs), into actionable insights for improving decision-making in healthcare and beyond. I explore broader questions of how both humans and machines understand and utilize language, aiming to develop systems that not only enhance decision-making across various domains but also contribute to a future where artificial intelligence (AI) is safe, trustworthy, and reliably aligned with human needs.