Deploying Apple’s machine learning tools on the server-side offers significant advantages.
During a recent AI Hackathon at Sovrn, participants were encouraged to explore machine learning concepts. The event showcased impressive projects, from SQL query generators to chatbots answering product-related questions. This presented an excellent opportunity to delve into Apple’s ML tools and potentially develop a solution with practical business value.
A team collaborated to experiment with CreateML and CoreML, aiming to integrate machine learning functionality into an iOS application. A model was successfully trained and integrated within hours, a notable achievement. However, several challenges became apparent that needed addressing before deployment.
- The model’s size was substantial, approximately 50MB, which occupies significant space within an app bundle.
- The desire was to update the model independently of new app version releases.
- There was also a need to utilize the model within a web browser environment.
Solving all these issues was not feasible at the time. However, during an exploration of the Vapor web framework, the idea emerged: why not deploy CoreML models on the server?
