Shift Technology runs an AI platform processing underwriting, claims, and fraud detection for 115+ insurers across 25 countries - including 6 of the top 10 US P&C carriers and 5 of the top 10 global insurers. The threat model here is mostly adversarial: organized fraud rings exploiting claims workflows, synthetic identity attacks, and insider schemes that scale faster than manual review teams can catch. The company fields over 200 insurance-focused data scientists building detection models that combine predictive, generative, and agentic AI to surface anomalies in real time across policy lifecycle events.
The technical stack runs on Azure with Kubernetes orchestration, using C# and ASP.NET on the backend with MSSQL and Elasticsearch for data layers. Frontend teams work in React, Redux, and TypeScript. CI/CD moves through GitHub Actions and TeamCity with Octopus handling deployments. Cypress covers end-to-end testing. The platform ingests structured claims data, unstructured documents, and third-party signals to score risk and flag patterns - accuracy matters because false positives block legitimate payouts and false negatives let fraud through.
Operations span nine cities: Paris (headquarters), Boston, Tokyo, Singapore, London, Madrid, Toronto, Mexico City, and São Paulo. The insurance domain creates specific challenges - regulatory variance across jurisdictions, legacy system integrations, and the need to explain model decisions to investigators and auditors. The work involves tuning models against adversarial adaptation, maintaining explainability under compliance requirements, and hardening data pipelines that handle sensitive PII at scale.
