Heidi built production-ready AI for healthcare at global scale

Australian AI care partner Heidi has successfully deployed production-ready AI solutions at a global scale for the healthcare industry. The company's flagship product, Heidi Scribe, now automates a significant portion of the administrative work consuming clinicians' time across more than 190 countries, supporting approximately 2.7 million patient interactions weekly.

According to Yu Liu, co-founder and CTO at Heidi, deploying AI in healthcare requires a different architectural approach compared to other sectors. While a small error rate might be an inconvenience elsewhere, the same rate in healthcare can pose a clinical safety issue. Liu emphasized that the architecture must be built with the assumption that every output will be scrutinized, audited, and relied upon in patient care.

Data residency and auditability are critical prerequisites for Heidi. Adhering to regulations like GDPR and HIPAA necessitates regional data storage. Heidi achieves this by maintaining logically isolated production deployments worldwide, ensuring patient data remains within its designated jurisdiction. Comprehensive audit trails are integrated from the outset to allow for retrospective analysis of data usage and outputs.

To manage the diverse and evolving medical data, Heidi utilizes MongoDB's document database. This provides the flexibility needed for rapidly changing AI data without constant database reshaping. Liu highlights that data architecture is crucial for system performance, and MongoDB's platform facilitates efficient handling of various data types, including transcripts and clinical notes. Heidi also leverages MongoDB Atlas's built-in vector search capabilities for semantic search, eliminating the need for separate vector databases.