Case Study: Humn.ai achieves real-time fleet risk scoring and faster production deployment with Ververica Platform

Machine Learning-based modelsto quantify commercial fleet exposureat individual insured asset risk in real-time

Humn.ai, an InsurTech startup building real-time, ML-driven fleet risk and pricing (Rideshur), faced limits with its initial streaming stack: stateful computations required extra lightweight datastores, and running Flink on AWS EMR added operational complexity that hindered a Kubernetes-native operational model. To address these challenges they turned to Ververica and Apache Flink, deploying the Ververica Platform to power their real-time risk scoring and dynamic pricing.

Using Ververica Platform with Apache Flink, Humn.ai moved to a Kubernetes-native production environment that continuously ingests vehicle and contextual data to produce per-asset risk scores and live premiums. Ververica helped Humn.ai productionize Flink jobs significantly faster, free developers to focus on features rather than deployment troubleshooting, and deliver smoother day-to-day operations and faster time-to-production.

Humn.ai

Alberto Romero

Co-founder and CTO