Serverless architecture for high-performance financial modelling

"fourTheorem have been a highly valued innovation partner for us. Super smart people with an in-depth knowledge of AWS. Their high calibre people are adept at applying the appropriate technology to solve complex business problems. We don't see enough of that in the market. An ability to attract that level of talent is a real differentiator."


RenaissanceRe is a global reinsurance business and has been in business for 29 years. With 11 global offices across EMEA, APAC and the Americas. In 2021 gross premiums were $8B with around $18B of capital under management. The key differentiator compared to the competition is their ability to use leading-edge technology to help them analyse risk and understand the portfolio better than anyone else.

The Challenge

The core process driving the understanding of a portfolio is the ‘Risk Rollup’, a model to present a view of risk. The process simulates events (e.g. earthquakes in the case of property reinsurance) to guide critical business decisions.

The system was running on-premises, and because of strong business growth, was under considerable stress. Specifically:

  • At current volumes, the system could run only 2-3 runs of portfolio rollups per day, each of which could take up to 10 hours.
  • They executed deal analytics in the same cluster, causing resource contention.
  • There was a growing requirement to run multiple, isolated, workloads to model additional market exposure scenarios.
  • The overall total cost of ownership (TCO) was high.

fourTheorem Solution

fourTheorem partnered with RenaissanceRe to deliver a full migration of the rollup process to AWS.

  • This involved re-architecture to leverage serverless & AWS managed services.
  • We began with discovery, design, & a six-week POC, delivering a working rollup that ran in 1 hour.
  • This POC was developed further to integrate the process into upstream & downstream processes.
  • The process was highly scalable, leveraging containers & Lambda functions to run batch processing & on-demand, real-time analytics.

The Outcome

  • Full portfolio rollups are complete in about one hour. The team knows that this can be reduced further by increasing the cluster size and adjusting the partitioning of modelling jobs. We can run multiple ad hoc portfolio rollup scenarios (‘what if’ analyses) within known time and cost parameters.
  • The overall codebase has been reduced by about 70% – lowering the total cost of ownership and removing responsibility for non-differentiating layers of the stack.
  • Analysts can run multiple rollups per day and concurrent rollups without having to negotiate for cluster time with colleagues. Previously, analysts were limited to weekends to run ad-hoc rollups for what-if analysis.
  • The breadth of compute services on AWS enables further cost optimisation, including the option of using Spot Capacity Providers on ECS.
  • New business capabilities are now possible with the ECS-based solution. RenaissanceRe has begun to run more advanced rollups with a 5-10x increase in data volume.

Next Steps

fourTheorem has followed this workload with a larger migration programme. This programme of workloads including, but not limited to:

  • Extended rollups with a 5-10x increase in data volume for a specific, novel business case.
  • Data lake reporting and aggregation using AWS Glue
  • Model development and publication using ECS with EFS
  • Platform development to facilitate inter-domain communication and data exchange, utilising EventBridge, AWS Glue and S3.