This course is customized for 1st-year PhD and MA students in the biomedical informatics graduate program and also
open to other interested students at Columbia. It provides a detailed overview of symbolic methods.
This course integrates the principles of database systems, AI agents, and health informatics. Students will gain expertise in the following areas:
Multimodal Data Engineering:
Ingesting, cleaning, and normalizing heterogeneous data (time-series sensor data, unstructured clinical notes, structured FHIR records)
Agentic Architectures:
Designing LLM-based agents capable of tool use (function calling), planning, and memory management (LangChain, LlamaIndex).
Retrieval-Augmented Generation (RAG):
Implementing vector databases and semantic search strategies optimized for personal health contexts.
Privacy and Ethics:
Understanding HIPAA compliance, differential privacy, and the ethical implications of AI "hallucinations" in health advice.
HCI for Health:
Designing applications and interfaces that foster trust, explainability, and engagement in self-tracking and patient-provider communication.
Domain Knowledge:
Understanding the basics of chronic disease management, personal health monitoring, and clinical decision support.
Prerequisite: open to public. Presentations by medical informatics faculty and invited international speakers in medical informatics, computer science, nursing informatics, library science, and related fields.
Research in medical informatics under the direction of a faculty adviser.
Research in medical informatics under the direction of a faculty adviser.
Prerequisite: instructors permission. Participation in medical informatics educational activities under the direction of a faculty adviser.
Prerequisite: completion of all M.Phil. requirements, and approval of a research proposal by the supervising faculty adviser.
Prerequisite: completion of all M.Phil. requirements. Ph.D. candidates may be required to register for this course every term during the preparation of the dissertation.