Private Health Management (PHM) specializes in care management for highly acute cancer. Working with self-insured companies, PHM’s team aimed to “clear cancer” by leveraging a machine learning model that predicts cancer in certain populations, helping employers understand their “cancer burden.” PHM also offers personalized assistance for clients’ employees, facilitating connections to optimal care and treatment solutions.
PHM needed a machine learning model that could scale with their company, analyzing employee workforce data from across the country to accurately predict the occurrence of 26 discrete cancers within unique populations.
PHM’s existing ML model was written off of R and had to be run and executed locally, requiring any user to download and configure the R programming language before executing the model in the command line in order to run it.
Hakkoda worked with PHM to clean up their provider data and craft a roadmap and analytics architecture that incorporated the organization’s key strategic initiatives and goals. This work removed room for human error around not only updating current versions of the model but distributing the model and all past and present data to any stakeholders
Working from the joint roadmap, Hakkoda’s team helped PHM build and migrate to a modern stack, identifying the best tooling and solution architecture for their goals. To scale their business, PHM leveraged a technology stack that included Snowflake, Fivetran, dbt, and Sigma. In addition to facilitating the onboarding, enablement, and migration to the modern data stack, Hakkoda worked with PHM to build a best-in-class data governance and quality program.
Hakkoda’s engineers operationalized PHM’s MLOps and Data Apps programs, eliminating the R programming that created scalability issues and migrating all projects and models to highly scalable, quality assured functionality on the modern data stack. This model reduces open source models completely and secures data and models within Snowflake’s secure environment. It also increases speed to delivery as all models run on the cloud rather than on local machines.