Philip J. Freda, Jr., Ph.D.

I am a Research Scientist I and clinical informatician in the Department of Computational Biomedicine at Cedars-Sinai Medical Center in Los Angeles, CA.

I develop AI-driven methods that make clinical data more usable for prediction, discovery, and decision support. My work sits at the intersection of clinical informatics, machine learning, and translational research, with a focus on building automated (and increasingly agentic) pipelines that transform noisy EHR data into clinically actionable representations. A major theme of my research is improving risk identification for adverse spine surgery outcomes and for problematic opioid use/opioid use disorder, leveraging structured EHR elements, clinical narratives, imaging, and social determinants of health to model patient risk and heterogeneity. 

I also design knowledge-graph–based frameworks that capture relationships among clinical concepts, data quality operations, and model behavior to support transparent, auditable AI workflows and human–AI collaboration. My earlier training in computational genetics and complex trait analysis informs how I think about high-dimensional interactions, robustness, and generalization, but my current emphasis is on clinically grounded, scalable AI systems for real-world healthcare data.

My main research focuses are:

Clinical informatics & AI for real-world data

• Agentic AI and automation for EHR data processing (data cleaning, feature engineering, reproducible pipelines)
• Phenotyping and risk modeling from structured + unstructured EHR data (including NLP)
• Clinical prediction for perioperative risk and outcomes in elective spine surgery 

Substance use disorders & clinical NLP

• Computational phenotyping of problematic opioid use / OUD from clinical notes and discharge summaries
• Severity characterization and context-aware NLP/annotation frameworks to support downstream modeling 

Knowledge graphs & AI-human collaboration

• Knowledge graph development to represent clinical concepts, data transformations, and model evidence
• Retrieval-augmented and graph-informed interfaces to improve interpretability, traceability, and actionability

Patient heterogeneity

• Quantifying patient subgroup structure and outcome heterogeneity (e.g., clustering approaches) in surgical populations
• Integrating and operationalizing social determinants data in EHR systems 

Computational genetics

• Complex trait modeling and interaction-focused methods (epistasis, non-additivity, AutoML for trait discovery)

If you are interested in my research or collaboration, have a look at my current research areas, recent publications, and/or email me!