20% reduction in deployment time by operationalizing risk scoring models for Largest P&C Insurance company

Problems:

• Rewrite the current financial risk scoring models from legacy SAS code to optimized Python script
• Build the end-to-end model deployment pipeline on MS Azure to enable both real-time and batch scoring capabilities
• Create Model APIs to allow integration with any front-end solution and build a model that can scale based on volume of request

Solution:

• Optimized and parametrized python models to allow real-time batch scoring
• 99.7% effective pass rate
• API based integration with all data sources required to feed the model to calculate the risk scores
• Containerized deployment of Model APIs, to provide the ability to scale on demand in real-time