F100 Retail Bank


Objective:
Speed up the process of deploying predictive models on to production systems.

 

Results Achieved: Created Data-as-a-Service platform called CMDS (Common Model Deployment System) on top of Hadoop to enable predictive models to move from Data Science lab to production in a fraction of the time it took earlier. The first model deployed on production reduced the wait period for a non-payment of loan notification from 20 days to just 1day.

 

  • Centralized risk scoring for bank customers across credit cards, auto and home loans
  • Data-as-a-Service platform via REST
  • Data lake ingestion from source systems
  • Data transformation & standardization
  • Monthly, daily and CDC from mortgage, auto, credit lending 
  • End-to-end model versioning and context made available
  • Data security added to the platform

 

Retail Bank

F100 Retail Bank

 

Objective: Enable graph analytics as self-service     

 

Results achieved: Created a platform to connect the customer and product hierarchies stored in a Graph Database to transactions from RDBMS and enabled roll-up, drill-down and slicing/dicing of data through a simple user-interface for business users to perform analytics by themselves. 

 

Using Neo4j, Talend, Groovy, Java, Cypher, SQL, Excel, CSV and JDBC  

  • Converted relational data model to graph
  • Designed and developed 9 use cases for different products
  • Validated that graph data model supports all 9 use cases
  • Migrated data in bulk loads
  • Measured and reported query performance KPIs