RADAR — Pipeline-Strike Prediction
A machine-learning model predicting third-party pipeline strikes from 811 tickets and GIS data — outperforming a $1M/yr vendor at under $100K and spawning a new Damage Prevention department.
Machine Learning Public Safety GIS MLflow
Public-safety ML
Third-party strikes on buried gas pipelines are a safety and cost risk. A vendor sold a prediction service for $1M/yr.
What I built
RADAR — a pipeline-strike prediction model in Python (scikit-learn, MLflow) built on 811 locate tickets and GIS spatial data, deployed on the company’s first cloud ML environment (Azure Databricks).
Impact
- Outperformed the $1M/yr vendor at under $100K
- Led directly to the creation of a new Damage Prevention department
- Owned end-to-end: feature engineering, training, deployment, monitoring, retraining