Abstract: Techniques for de-centralized execution of workflows are disclosed. A system executes a sequence of jobs in a workflow. Each job includes application code to perform a defined set of functions. Each job includes wrapper code at one or both ends of the application code. The system executes the sequence of jobs by iteratively spinning up a virtual machine, loading workflow data to be used by application code, executing the application code, and storing modified workflow data. The virtual machine executing one job triggers the next virtual machine to execute the next job in the workflow, without scheduling the next job by a centralized job scheduler. Upon spinning up the next virtual machine, a virtual machine executing a preceding job shuts itself down.

Abstract: The present disclosure relates generally to an integrated machine learning platform. The machine learning platform can convert machine learning models with different schemas into machine learning models that share a common schema, organize the machine learning models into model groups based on certain criteria, and perform pre-deployment evaluation of the machine learning models. The machine learning models in a model group can be evaluated or used individually or as a group. The machine learning platform can be used to deploy a model group and a selector in a production environment, and the selector may learn to dynamically select the model(s) from the model group in the production environment in different contexts or for different input data, based on a score determined using certain scoring metrics, such as certain business goals.

Abstract: The present disclosure relates generally to an integrated machine learning platform. The machine learning platform can convert machine learning models with different schemas into machine learning models that share a common schema, organize the machine learning models into model groups based on certain criteria, and perform pre-deployment evaluation of the machine learning models. The machine learning models in a model group can be evaluated or used individually or as a group. The machine learning platform can be used to deploy a model group and a selector in a production environment, and the selector may learn to dynamically select the model(s) from the model group in the production environment in different contexts or for different input data, based on a score determined using certain scoring metrics, such as certain business goals.

Abstract: The present disclosure relates generally to an integrated machine learning platform. The machine learning platform can convert machine learning models with different schemas into machine learning models that share a common schema, organize the machine learning models into model groups based on certain criteria, and perform pre-deployment evaluation of the machine learning models. The machine learning models in a model group can be evaluated or used individually or as a group. The machine learning platform can be used to deploy a model group and a selector in a production environment, and the selector may learn to dynamically select the model(s) from the model group in the production environment in different contexts or for different input data, based on a score determined using certain scoring metrics, such as certain business goals.

Abstract: A real-time search system enables askers to identify and submit questions to topically and skill level relevant potential answerers. A computer server receives and analyzes short text questions, determines a corresponding set of informational facets semantically and topically characterizing the question. The informational facets are evaluated against a database index of informational facets identified from prior analyzed messages correlated by profile identifiers of message originators to provide an identification of a plurality of potential answerers. The question is distributed to the plurality of potential answerers and ensuing message conversations between the asker responsive answerers are monitored for quality and sufficiency of response. The stored profiles of responsive answerers are updated to reflect the occurrence and quality of response.

Abstract: We disclose techniques for varying the caching of content provided by a content server as a function of the server's load. When the server is lightly loaded, freshness of the content is maintained. As server load increases, caching time increases, to trade off freshness against transmission time. Similarly, when the server is heavily loaded, users might quickly be served cached content that is only slightly stale—as opposed to content that was fresh at the time of the request, but which becomes materially stale by the time the server completes serving the response. The server's load can be measured by its response time to a request, or otherwise. Optionally, the system can override the load-based caching by defining classes of requests that automatically trigger updating or refreshing (e.g., cache expiration) of related information in the cache.

Abstract: We disclose techniques for varying the caching of content provided by a content server as a function of the server's load. When the server is lightly loaded, freshness of the content is maintained. As server load increases, caching time increases, to trade off freshness against transmission time. Similarly, when the server is heavily loaded, users might quickly be served cached content that is only slightly stale—as opposed to content that was fresh at the time of the request, but which becomes materially stale by the time the server completes serving the response. The server's load can be measured by its response time to a request, or otherwise. Optionally, the system can override the load-based caching by defining classes of requests that automatically trigger updating or refreshing (e.g., cache expiration) of related information in the cache.