Sempre Health Case Study: Accelerating NLP Pipelines with Hugging Face Expert Acceleration Program

Sempre Health collaborated with Hugging Face through the Expert Acceleration Program to implement a machine learning pipeline for automated message classification. This partnership enabled Sempre Health to automate the handling of over 20% of incoming patient text messages, reducing low-value manual work for their operations team and increasing overall business scalability.

Automated Patient Message Classification

Sempre Health uses Natural Language Processing (NLP) to identify and respond to thousands of daily text messages from patients regarding medication adherence and affordability. The system is designed to handle common requests, such as simple acknowledgments ("Thank you") or specific service requests like prescription refills, which trigger automated pharmacy notifications.

Transition from Rule-Based to Statistical ML

Before adopting the Expert Acceleration Program, Sempre Health relied on a rule-based system that successfully captured approximately 80% of inbound messages. To improve parsing accuracy and efficiency, the company transitioned to a statistical machine learning approach using language models from Hugging Face.

Despite having internal expertise in ML and NLP, the team faced uncertainty regarding:

  • The optimal formulation of the problem.
  • The selection of the most effective model or neural network architecture for their specific use case.
  • The most efficient way to utilize their available training data.

Role of the Expert Acceleration Program

The Expert Acceleration Program provided technical guidance across the entire implementation lifecycle of the NLP solution. Key contributions from the Hugging Face team included:

  • Data Labeling: Providing advice on obtaining representative and accurate labels for text messages.
  • Model Selection: Reducing research time by identifying the appropriate models and methods immediately.
  • Development Speed: Expediting the development process, allowing the team to reach production faster than they would have independently.

Business Impact and Scalability

Since launching the production NLP pipeline, Sempre Health has consistently seen nearly 20% of incoming messages handled automatically by the new system. These messages previously required the creation of a manual ticket for the patient operations team, meaning the automation has directly reduced the volume of low-value manual tasks and improved operational workflow.

Recommendations for ML Teams

Sempre Health recommends the Expert Acceleration Program for small teams or organizations looking to integrate AI pipelines into their products quickly. The program is specifically suited for teams aiming to minimize the development time and research hassle typically associated with machine learning development.

Sources