Case Study: Appen achieves scalable fraud detection with Provectus

Appen efficiently detects and prevents scam with a scalable, ML-powered fraud detection platform

Appen, a leading provider of training data for AI systems, needed to replace a mostly manual fraud detection process that was triggered by SQL and Python scripts. As its platform grew, Appen struggled to monitor more than 50 jobs per day efficiently and wanted a scalable way to detect malicious activity while reducing manual work and human error. Provectus helped by designing an automated fraud detection platform with human-in-the-loop capabilities, using TensorFlow and AWS-based infrastructure.

Provectus built and integrated end-to-end data pipelines, trained and tuned ML/DL models, and developed a user-friendly web application to manage fraud alerts and moderation workflows. The solution, deployed with Hydrosphere.io on Amazon ECS and supported by Prometheus and Grafana monitoring, enabled Appen to handle 20x more jobs per day with 97% automated processing, reduce scammers by 25%, and achieve 5x fewer churned judgments, while avoiding the need to hire 20+ data analysts.

Appen