Flo Health
A developer of a leading global women's health tracking application, offering health management services covering menstrual cycles, ovulation and pregnancy, with a large user base.
Timeline (last 90 days)
Flo Health is one of the most popular women's health apps in the world, with monthly active users as high as 81 million.
Flo Health does not directly call closed-source large model APIs, but instead has built a more controllable and safer open model system based on Databricks AI Runtime.
Flo Health mixes high-risk examples such as self-harm and emergency medical conditions, boundary cases annotated by clinical experts, and instruction data for "not giving diagnostic conclusions" into the fine-tuning set, so that the model learns at the parameter level to refuse or refer.
On Databricks AI Runtime, Flo Health's safety strategy includes input filtering, output constraints, a human evaluation loop, and observability, with safety capabilities internalized into the fine-tuning data and evaluation process.
In Flo Health's practice, Databricks AI Runtime plays key roles such as unified compute scheduling, inference performance optimization, versioning and rollback, and evaluation pipelines.
Flo Health's path is not "training from scratch on its own", but rather domain adaptation based on open-source foundation models.
Flo Health explicitly states that when building AI features, "safe" and "useful" are not parallel—safety is a precondition.
All Facts (7)
Flo Health is one of the most popular women's health apps in the world, with monthly active users as high as 81 million.
50%UnverifiedFlo Health does not directly call closed-source large model APIs, but instead has built a more controllable and safer open model system based on Databricks AI Runtime.
50%UnverifiedFlo Health mixes high-risk examples such as self-harm and emergency medical conditions, boundary cases annotated by clinical experts, and instruction data for "not giving diagnostic conclusions" into the fine-tuning set, so that the model learns at the parameter level to refuse or refer.
50%UnverifiedOn Databricks AI Runtime, Flo Health's safety strategy includes input filtering, output constraints, a human evaluation loop, and observability, with safety capabilities internalized into the fine-tuning data and evaluation process.
50%UnverifiedIn Flo Health's practice, Databricks AI Runtime plays key roles such as unified compute scheduling, inference performance optimization, versioning and rollback, and evaluation pipelines.
50%UnverifiedFlo Health's path is not "training from scratch on its own", but rather domain adaptation based on open-source foundation models.
50%UnverifiedFlo Health explicitly states that when building AI features, "safe" and "useful" are not parallel—safety is a precondition.
50%