An insurer wanted to use data science for advanced actuarial modelling and predict risk based on individual behaviours. However, data sat in legacy systems and required huge manual effort to build models.
UpSky Tech (UST) offered to migrate data to AWS cloud data warehouse and lake built using Python scripts; Data pipeline automation, Spark workloads and SageMaker machine learning delivered real-time risk insights.
Technical Details:
- Automated data ingestion via AWS Glue ETL jobs
- Structured customer data stored in AWS Redshift
- Unstructured sensor data lake running on S3
- Batch data analytics pipeline with Apache Spark
- Jupyter notebooks on SageMaker to develop ML models in Python leveraging algorithms like random forest and logistic regression
- Performance testing on load generation service, AWS X-Ray for tracing
- API data access using AWS Lambda functions for integration with backend systems
- Dashboards and data visualization built using AWS Quicksight
- Risk models accurate to 98%, reducing errors
- Claims analysis done in hours instead of weeks
- Underwriters more effective with customer insights