Abstract: A probabilistic loss function based machine learning system (MLS) is disclosed. The MLS can generate an additional new class for uncertainty value(s) determined for inputs to the MLS. The uncertainty value(s) can be correlated to output(s) of the MLS. A novel loss function can be based on a conventional loss function, the uncertainty value(s), and an adjustable penalty value. The adjustable penalty value can be adjusted, such that improving the optimization of the novel loss function can cause the MLS to reduce use of computing resources in determining outputs corresponding to uncertainty value(s) above a threshold value(s) in favor of using computing resource for determining outputs corresponding to uncertainty value(s) below the threshold value(s). Updating of the adjustable penalty value can result in changes to the threshold value(s).

Abstract: In one aspect, an example methodology implementing the disclosed techniques includes, in response to receipt at a call center of a call by a caller, generating a compatibility score for the caller, the compatibility score for the caller optimizes for caller satisfaction and potential upsell opportunities. The method also includes generating a compatibility score for individual agents available to handle the call, the compatibility score for the individual agents optimize for caller satisfaction and potential upsell opportunities. The method also includes matching the compatibility score for the caller to the compatibility scores for the individual agents and assigning an agent to the call based on the matching of the compatibility scores for the caller and the individual agents.

Abstract: In some examples, a server may receive usage data from a computing device, determine a usage profile of the computing device, and determine that a component of the computing device is predicted to fail and at what time. If the time is within a warranty period of a current warranty, then the server may recommend purchasing an upgraded warranty and provide a cost-benefit analysis of purchasing the upgraded warranty. If the time is outside the warranty period, then the server may recommend purchasing an extended warranty and provide a cost-benefit analysis of purchasing the extended warranty. The server may use the usage data to make a recommendation to upgrade one or more components (e.g., memory, disk drive, network card, or the like) of the computing device. The server may use the usage data to make a recommendation to upgrade from the computing device to a different computing device.

Abstract: In one aspect, an example methodology implementing the disclosed techniques includes, in response to receipt at a call center of a call by a caller, generating a compatibility score for the caller, the compatibility score for the caller optimizes for caller satisfaction and potential upsell opportunities. The method also includes generating a compatibility score for individual agents available to handle the call, the compatibility score for the individual agents optimize for caller satisfaction and potential upsell opportunities. The method also includes matching the compatibility score for the caller to the compatibility scores for the individual agents and assigning an agent to the call based on the matching of the compatibility scores for the caller and the individual agents.

Abstract: In some examples, a server may receive a user communication describing an issue with a computing device and assign a case to the computing device. The server may determine previously provided telemetry data (e.g., logs and usage data sent by the computing device) as well as previous cases associated with the computing device. Machine learning may be used to predict, based on the user communication, the telemetry data, and the previous cases, a predicted cause of the issue, a predicted time to close the case, and a predicted set of steps to resolve the issue. The machine learning may predict a bottleneck in at least one step of the set of steps that causes the predicted time to close to exceed a threshold and predict one or more actions to address the bottleneck. The server may automatically perform at least one action of the one or more actions.

Abstract: In some examples, a server may receive usage data from a computing device, determine a usage profile of the computing device, and determine that a component of the computing device is predicted to fail and at what time. If the time is within a warranty period of a current warranty, then the server may recommend purchasing an upgraded warranty and provide a cost-benefit analysis of purchasing the upgraded warranty. If the time is outside the warranty period, then the server may recommend purchasing an extended warranty and provide a cost-benefit analysis of purchasing the extended warranty. The server may use the usage data to make a recommendation to upgrade one or more components (e.g., memory, disk drive, network card, or the like) of the computing device. The server may use the usage data to make a recommendation to upgrade from the computing device to a different computing device.