Outcome: Built an interpretable risk-modeling framework to support occupational health decision-making when direct exposure data are limited, emphasizing early risk detection and conservative safety thresholds.

Decision Context

• Workplace exposure risks are difficult to monitor directly in many operational environments
• Decision-makers still need tools to prioritize monitoring and prevention efforts
• False negatives carry higher cost than false positives in safety-critical settings

What I Built

• Simulation-based cohort: Generated structured synthetic data grounded in published exposure research to enable scenario analysis
• Predictive modeling pipeline: Trained and compared multiple tree-based classifiers for nonlinear risk relationships
• Safety-oriented evaluation: Tuned thresholds to prioritize recall and minimize missed high-risk cases
• Interpretability layer: Applied feature attribution to support transparent review and communication of risk drivers
• Reproducible workflow: Delivered an end-to-end pipeline from data generation through evaluation

Impact

• Demonstrated a practical approach for decision support when real-world data are incomplete
• Provided interpretable signals to guide monitoring and prevention prioritization
• Established a reusable framework adaptable to other low-observability risk domains

This work focuses on decision support and risk prioritization rather than clinical diagnosis. Simulation is used deliberately to explore feasibility and trade-offs under realistic constraints.