Company Background
The client is a publicly listed Australian utility company involved in the generation and retailing of electricity and gas for residential and commercial use. The company operates a significant generation capacity of 10,984 MW, with 85% of this capacity derived from coal-fired plants. This dual-role as both generator and retailer places the company in a unique position within the market, necessitating robust and efficient management of both operational and customer-facing activities.
The company faced multiple challenges impacting its operations and customer service:
Debt Management
Data Management
Operational Efficiency
Machine Learning Integration
To address these challenges, a comprehensive solution was implemented leveraging advanced data management and machine learning technologies:
Created a centralised data lake using Azure Storage and Databricks to aggregate data from multiple source systems.
Implemented a Spark streaming framework to handle data ingestion, ensuring real-time data availability for critical operational reports.
Organised data into different layers (staging, raw, integrated, and transformed) for efficient processing and retrieval.
Established an enterprise ML platform with a feature store accessible to all ML engineers and data scientists, minimising feature duplication and fostering collaboration.
Utilised machine learning algorithms to predict customers with a high propensity to fall into debt, based on transactional and behavioral data from the ERP system.
The solution was hosted within the company’s environment to comply with stringent data privacy policies.
Conducted thorough causal analysis to understand the underlying factors leading to debt, enabling targeted interventions.
Formed specialist teams to investigate and address potential issues identified by the AI/ML models, such as billing errors or faulty meters.
Leveraged Databricks' optimisation features such as adaptive query execution and dynamic partition pruning to enhance processing speed and resource utilisation.
Implemented centralised data governance using Databricks Unity Catalog and Collibra to ensure compliance and data integrity.
Developed end-to-end MLOps pipelines using Databricks and Azure DevOps to automate model deployment and monitoring.
The implementation of these solutions yielded significant business outcomes:
Debt Reduction
Revenue Realisation