ML-Driven Approval Prediction for Dental Prosthesis Claims

Manual review of dental prosthesis prescriptions had become a bottleneck for the client. High case volumes led to long processing times and inconsistent decisions across reviewers, while valuable predictive signals already present in the application datasets remained untapped.

Templefield Technologies was engaged to build a scalable ML platform that predicts the approval probability of dental prosthesis claims in real time — designed from the ground up to satisfy strict data-privacy standards and EU AI Act requirements.

• Built automated, anonymized data pipelines from the client's SAP systems into S3 using dlt
• Established a microservices architecture on Kubernetes with Champion/Challenger testing patterns for safe model iteration
• Delivered a real-time REST inference service with automated model lifecycle management
• Integrated an MLflow model registry with encrypted storage and continuous compliance monitoring
• Deployed the platform via Helm using a GitOps workflow for reproducible, auditable releases

This engagement is in progress, running January–December 2026. Templefield is building a compliant, real-time approval-prediction platform for multiple AOK health insurers, aiming to automate straightforward decisions, route complex cases to human reviewers, and improve decision consistency while embedding data protection and regulatory compliance by design.