Providing models to client devices

1. A system for providing an enterprise-specific neural network model to a client device, comprising:

at least one computing device; and

program instructions executable in the at least one computing device that, when executed by the at least one computing device, cause the at least one computing device to:

create a training environment for training a neural network using enterprise data authorized by a client device, the training environment being accessible by a locator;

cause the client device to send the enterprise data to the training environment by generating a request for the client device to authorize sending the enterprise data to the training environment, transmitting a request for the client device to authenticate with the training environment, causing a certificate to be created, the certificate being associated the client device, and causing the client device to authenticate with the training environment using the certificate and the locator;

cause the training environment to create a model for performing a text classification based on providing the enterprise data to an input associated with the training environment; and

send the model to the client device.

2. The system of claim 1 , wherein the program instructions create the training environment by at least:

instructing an infrastructure manager to create a cloned environment based on a template;

causing a hypervisor to start at least one virtual machine associated within the cloned environment, the at least one virtual machine configured to run a training application; and

wherein the input associated with the training environment comprises at least one parameter of the training application.

3. The system of claim 2 , wherein the training application comprises program instructions executable to cause a processor associated with the virtual machine to:

preprocess the enterprise data to create tokenized data;

execute a TensorFlow framework to create the model based on applying a Bag of Words approach to the tokenized data; and

store the model.

4. The system of claim 1 , wherein sending the model to the client device comprises:

causing a push notification to be sent to the client device.

5. The system of claim 1 , wherein the enterprise data is client data stored in a data store associated with the client device.

6. The system of claim 1 , further comprising:

wherein the model contains a topology and weights extractable using a basic neural network subroutines (BNNS) library, the topology relating at least one input node to at least one output node, the weights representing knowledge gained through training the neural network;

wherein the text classification comprises applying one or more emails to the at least one input node, the text classification outputting to the at least one output node; and

wherein the at least one output node represents a sentiment of the email.

7. A computer-implemented method for providing an enterprise-specific neural network model to a client device, comprising:

creating, by a computing device, a training environment for training a neural network using enterprise data authorized by a client device, the training environment being accessible by a locator;

causing, by the computing device, the client device to send the enterprise data to the training environment by generating a request for the client device to authorize sending the enterprise data to the training environment, transmitting a request for the client device to authenticate with the training environment, causing a certificate to be created, the certificate being associated the client device, and causing the client device to authenticate with the training environment using the certificate and the locator;

causing, by the computing device, the training environment to create a model for performing a text classification based at least in part on providing the enterprise data to an input associated with the training environment; and

sending, by the computing device, the model to the client device by causing the client device to authenticate with the training environment to receive the model.

8. The method of claim 7 , the creating a training environment further comprising:

instructing an infrastructure manager to create a cloned environment based on a template;

causing a hypervisor to start at least one virtual machine associated within the cloned environment, the at least one virtual machine configured to run a training application; and

wherein the input associated with the training environment comprises at least one parameter of the training application.

9. The method of claim 8 , the training application further comprising:

preprocessing the enterprise data to create tokenized data;

executing a TensorFlow framework to create the model based on applying a Bag of Words approach to the tokenized data; and

storing the model.

10. The method of claim 7 , wherein sending the model to the client device comprises:

causing a push notification to be sent to the client device.

11. The method of claim 7 , wherein the enterprise data is client data stored in a data store associated with the client device.

12. The method of claim 7 , further comprising:

wherein the model contains a topology and weights extractable using a BNNS library, the topology relating at least one input node to at least one output node, the weights representing knowledge gained through training the neural network;

wherein the text classification comprises applying one or more emails to the at least one input node, the text classification outputting to the at least one output node; and

wherein the at least one output node represents a sentiment of the email.

13. A non-transitory computer-readable medium for providing an enterprise-specific neural network model to a client device embodying program instructions executable in a computing device that, when executed, causes the computing device to:

create a training environment for training a neural network using enterprise data authorized by a client device, the training environment being accessible by a locator;

cause the client device to send the enterprise data to the training environment by generating a request for the client device to authorize sending the enterprise data to the training environment, transmitting a request for the client device to authenticate with the training environment, causing a certificate to be created, the certificate being associated the client device, and causing the client device to authenticate with the training environment using the certificate and the locator;

cause the training environment to create a model for performing a text classification based at least in part on providing the enterprise data to an input associated with the training environment; and

send the model to the client device by causing the client device to authenticate with the training environment to receive the model.

14. The non-transitory computer-readable medium of claim 13 , wherein executing the program instructions further causes the training environment to be created by:

instructing an infrastructure manager to create a cloned environment based on a template;

causing a hypervisor to start at least one virtual machine associated within the cloned environment, the at least one virtual machine configured to run a training application; and

wherein the input associated with the training environment comprises at least one parameter of the training application.

15. The non-transitory computer-readable medium of claim 14 , wherein the training application comprises program instructions executable to cause a processor associated with the virtual machine to:

preprocess the enterprise data to create tokenized data;

execute a TensorFlow framework to create the model based on applying a Bag of Words approach to the tokenized data; and

store the model.

16. The non-transitory computer-readable medium of claim 13 , wherein the enterprise data is client data stored in a data store associated with the client device.

17. The non-transitory computer-readable medium of claim 13 further comprising:

wherein the model contains a topology and weights extractable using a basic neural network subroutines (BNNS) library, the topology relating at least one input node to at least one output node, the weights representing knowledge gained through training the neural network;

wherein the text classification comprises applying one or more emails to the at least one input node, the text classification outputting to the at least one output node; and

wherein the at least one output node represents a sentiment of the email.