Machine Learning Engineer
CurrentMachine Learning / Predictive Modeling:• Created a Forecasting DeepAR Model to predict Water demand for multiple Connections and was able to bring down the run time of batch model processing.• Designed a XGBoost classification model on imbalanced time series data to predict the connection category and achieved 75% recall which helped the utility increase revenue.• Built a leak clustering DBSCAN model to localize leak hotspots in the current water network and did rule extraction on these hotspots using Decision Trees Classification model to help utilities find unidentified leaks in the Water Network.Data / Machine Learning Operations:• Setup DVC (Data Version Control) for versioning data on the cloud.• Setup great_expectations in production to do Data Quality Checks before data hits the Data Lake.• Setup self-managed Kubeflow on an existing Kubernetes cluster and productionized 6 models using Kubeflow Pipelines and Model Registry.Software Systems:• Created APIs for the Backend to integrate the results from ML Runs to the UI.• Created Multiple KPIs for the Water demand/supply and integrated them into the UI using APIs.
