Upgrading our inference engine

We're rolling out a faster model with improved accuracy. The platform will be briefly unavailable while we complete the deployment.