Data Scientist I - Windreich Department for AI & Human Health Research

The Data Scientist will join the PREVENT pAKI project (NHLBI-funded), which uses federated and pooled machine learning on EHR data from 15 hospitals and more than 320,000 non-cardiac surgery patients. The aims are to predict postoperative acute kidney injury (pAKI) and to learn personalized prevention strategies that generalize across patient subgroups. You will work directly with the PI, a nephrologist co-investigator, and site teams at Mount Sinai, UCLA and Vanderbilt. The role will leverage and create cutting edge machine learning techniques and emphasize statistical rigor, reproducibility and scientific publication.

Ideal candidate is a Master’s graduate who will use their 2+ years of background experience in healthcare data science/analytics to successfully build cohorts of pAKI patients and develop deep learning models to predict pAKI risk in future patients. This candidate is comfortable with independently performing statistical analyses, can communicate effectively and collaborate well with project stakeholders/peers. Within 1 year, candidate will have deployed a successful deep learning model using PyTorch and be able to demonstrate performance of outcome prediction suitably for technical and non-technical audiences.

• Build and curate analytic cohorts from OMOP-mapped and perioperative data warehouse sources, including AKI outcome definitions (AKIN/KDIGO) and quality checks for consistency, plausibility and missingness.

• Develop and validate prediction models (gradient-boosted trees, graph neural networks) with uncertainty quantification, and report discrimination, precision-recall and calibration.

• Implement and compare site-specific, pooled and federated (stacked) models across three institutions and various subgroups and implement mitigation strategies for underperforming groups.

• Apply causal inference methods (causal forests, policytree) and reinforcement learning to identify sub-phenotypes and personalized prevention strategies.

• Maintain version-controlled, reproducible code and documentation sufficient for external replication, and prepare code and models for public release.

• Contribute to manuscripts, abstracts and presentations, and communicate methods and results clearly to clinicians and non-technical audiences.

• Coordinate with site teams on data requests, local model training, and project milestones.

• Stay current with federated learning, causal ML and algorithmic fairness literature and propose methodological improvements.

• Other duties as assigned.

• Master's degree in a quantitative discipline (e.g., Statistics, Operations Research, Bioinformatics, Economics, Computational Biology, Computer Science, Information Technology, Mathematics, Physics) or equivalent practical experience.
• 2 years of work experience in data science, software engineering, or data analysis
• Experience with at least one programming language among Scala, Python, Java, C, or C++.
• Experience with database languages (e.g., SQL, NoSQL)
• Familiarity with cloud computing platforms (e.g., AWS, Azure, GCP)
• Experience with version control systems (e.g., Git)
• Knowledge of big data technologies (e.g., Hadoop, Spark)
• Self-motivated with a demonstrated ability to work independently, and to exercise independent judgment in developing complex techniques or programs in a dynamic environment

Preferred:

• Experience with EHR data (OMOP preferred) and clinical prediction modeling, including calibration and external validation.

• Causal inference or heterogeneous treatment effect methods (e.g., grf/policytree).

• Deep learning (PyTorch, graph neural networks), federated learning, or algorithmic fairness evaluation.

• Cloud or HPC experience (AWS, Azure, GCP); PhD; first-author publications.