Core ML represents a horizontal effort within our Machine Learning (ML) group to create libraries that are reusable across our solutions, to open source code and data, and to establish internal best practices for ML code and model development.
While our projects span various technical domains within ML, ranging from computer vision and audio signal processing to tabular data analysis, each of these domains contain reusable components that are likely to be useful across projects. For example, classification tasks on tabular data, along with pre-processing of noisy data, are likely to be ubiquitous for our work in health. Object detection tasks, measurement tasks on images and videos, creating fixed length dense representations for text and multimedia represent other areas of work that are likely to find multiple applications in the social domain.
As a nonprofit devoted to creating social impact through AI, we also support and practice open-sourcing of code and data wherever possible. Our core ML team is responsible for stewarding the open-sourcing of code that was primarily built for internal consumption within our solutions, ensuring that code is well documented and bug-free, and available for public use under the appropriate open source license(s). It is also responsible for ensuring that data that we share externally is stripped of any Personally Identifiable Information (PII) and that its availability is appropriately credentialed.