Catching invisible errors: How I built a duplicate detection agent for Kenya's HIV program

To combat the high rate of undetected duplicate patient records in Kenya’s HIV program, the author developed a multi-agent system using Elasticsearch and Claude Sonnet 3.7. The system utilizes three specialized agents—detection, risk assessment, and action recommendation—to rapidly identify, score, and provide actionable insights for potential duplicates. This automated approach enables healthcare workers to efficiently prioritize critical errors, preventing wasted medical resources and ensuring data integrity.