Scalable architecture for machine learning model reuse

A system comprising:

• at least one data processor;
• and
• memory storing instructions which, when executed by the at least one data processor, result in operations comprising:
• receiving a request for a recommendation for one or more machine learning models in a group of pre-determined, reusable machine learning models in a model repository, the respective pre-determined, reusable machine learning models including metadata that describes a respective defined schema, the request encapsulating parameters for a desired machine learning model, wherein the encapsulated parameters include one or more of an application, a request type, a technical stack, an input type, or a desired prediction;
• identifying, by a machine learning-based discovery model using the encapsulated parameters including one or more of the application, the request type, the technical stack, the input type, or the desired prediction, at least one recommended machine learning model, from the group of pre-determined, reusable machine learning models, that are usable to address the request, wherein an aggregator layer searches the model repository for common fields within the pre-determined, reusable machine learning models in the model repository and forms a link amongst multiple pre-determined, reusable machine learning models having common fields to combine the outputs from the multiple pre-determined, reusable machine learning models and form an aggregated output from such multiple pre-determined, reusable machine learning models, and wherein the aggregator layer is built using a template when an Application Programming Interface (API) corresponding to the at least one recommended pre-determined, reusable machine learning model is not already available;
• and
• providing data characterizing the at least one recommended pre-determined, reusable machine learning model.

A non-transitory computer program product storing instructions which, when executed by at least one computing device, result in operations comprising:

• receiving a request for a recommendation for one or more machine learning models in a group of pre-determined, reusable machine learning models in a model repository, the respective pre-determined, reusable machine learning models having a respective defined schema, the request encapsulating parameters for a desired machine learning model, the encapsulated parameters including one or more of an application, a request type, a technical stack, an input type, or a desired prediction;
• extracting features from the request and inputting the extracted features into a machine learning-based discovery model;
• identifying, by the machine learning-based discovery model using the encapsulated parameters including one or more of the application, the request type, the technical stack, the input type, or the desired prediction, at least one recommended pre-determined, reusable machine learning model corresponding to the request, wherein an aggregator layer searches the model repository for common fields within the schemas of the pre-determined, reusable machine learning models in the model repository and forms a link amongst multiple pre-determined, reusable machine learning models having the common fields to form an aggregated output from the multiple pre-determined, reusable machine learning models, and wherein the aggregator layer is built using a template when an Application Programming Interface (API) corresponding to the at least one recommended pre-determined, reusable machine learning model is not already available;
• and
• providing data characterizing one or more of the aggregated output or the at least one recommended pre-determined, reusable machine learning model.