Bayes Server Risk Collaboration

We collaborated with Bayes Server and a third party risk company in the UK to put together a solution for automatically generating new Bayesian risk frameworks based upon the third party's client data. The client required a better way of connecting leaf nodes in their own risk framework, typically metrics, and surfacing these in a probabilistic manner to the root nodes which were monitored for alerting purposes.

There was potential for large networks to be generated so the work included developing novel network structures to improve inference efficiency, as well as new ways to specify variable parameterisation based upon customer configuration. The client's own customers had the freedom to manage their own acceptable levels of risk, and variables contributing to the risk, so it was necessary to be able to rebuild the network promptly dependent upon changes to the underlying risk structure.

This was a C# backend project which hooked in to gRPC endpoints, and was hosted as part of the client's web application.