Integration of heterogeneous models into robotic process automation workflows

Patent metadata

Field	Value
Publication number	US-11738453-B2
Application number	US-201916710027-A
Country	US
Kind code	B2
Filing date	Dec 11, 2019
Priority date	Oct 15, 2019
Publication date	Aug 29, 2023
Grant date	Aug 29, 2023

Abstract

Frameworks and techniques for integration of heterogeneous machine learning (ML) models into robotic process automation (RPA) workflows are provided. This may be accomplished via a seamless drag-and-drop interface that allows deployment of ML models into an RPA workflow. Via a framework, these heterogeneous models may be provided by customers, third parties, and/or partners and integrated into the RPA workflow. The framework may provide a straightforward way to deploy machine learning models via a conductor and to manage model versioning and create/retrieve/update/delete (CRUD). The framework may facilitate integration of different models into the RPA workflow through the steps of uploading, validating, publishing, and deploying models.

First claim

The invention claimed is:

1. A non-transitory computer-readable medium storing a computer program, the computer program configured to cause at least one processor to:
receive a machine learning (ML) model from a conductor application;
perform validation on the ML model;
responsive to validation of the ML model succeeding:
upload the ML model into storage,
deploy the ML model for use by robotic process automation (RPA) robots, and
publish the ML model by exposing the ML model as a service via an Application Programming Interface (API) that the RPA robots call during execution; and

responsive to validation of the model failing:
reject the ML model.

2. The non-transitory computer-readable medium of claim 1 , where the computer program is further configured to cause the at least one processor to:
receive a request from an RPA robot to execute the deployed ML model;
receive data to be used by the deployed ML model;
execute the deployed ML model using the received data; and
transmit results of the executed ML model to the RPA robot.

3. The non-transitory computer-readable medium of claim 2 , wherein the deployed ML model is an initial version of the ML model.

4. The non-transitory computer-readable medium of claim 1 , wherein the ML model is received from the conductor application via a proxy that is loosely coupled with the conductor application, the proxy configured to be decoupled from the conductor application and run independently therefrom.

5. The non-transitory computer-readable medium of claim 4 , wherein the proxy is configured to tunnel requests associated with the ML model via an Application Programming Interface (API) gateway of a service tier.

6. The non-transitory computer-readable medium of claim 1 , wherein when the ML model is a new version of an existing ML model, the computer program is further configured to cause the at least one processor to:
perform version control on the ML model by creating and storing metadata regarding the new version of the ML model.

7. The non-transitory computer-readable medium of claim 1 , wherein the computer program is further configured to cause the at least one processor to:
upload, validate, and deploy the ML model via a service tier.

8. The non-transitory computer-readable medium of claim 7 , wherein the service tier is configured to provide internal utility services that perform asynchronous operations, state machine management, storage abstraction and access, or any combination thereof.

9. The non-transitory computer-readable medium of claim 7 , wherein the service tier is configured to provide a model publish service that performs create, retrieve, update, and delete (CRUD) operations.

10. The non-transitory computer-readable medium of claim 7 , wherein the service tier is configured to build images of the ML model with dependencies, build wrapper code around the ML model to create a container, push the container to a container registry, and deploy the container as an ML skill for consumption by the RPA robots.

11. The non-transitory computer-readable medium of claim 1 , wherein the API is a Representative State Transfer (REST) API.

12. A computer-implemented method, comprising:
receiving a machine learning (ML) model from a conductor application, by a computing system;
performing validation on a machine learning (ML) model, by the computing system; and
responsive to validation of the ML model succeeding:
uploading the ML model into storage, by the computing system,
deploying the ML model, by the computing system, for use by robotic process automation (RPA) robots, and
publishing the ML model by exposing the ML model as a service via an Application Programming Interface (API) that the RPA robots call during execution.

13. The computer-implemented method of claim 12 , further comprising:
receiving, by the computing system, a request from an RPA robot to execute the deployed ML model;
receiving data to be used by the deployed ML model, by the computing system;
executing the deployed ML model using the received data, by the computing system; and
transmitting results of the executed ML model to the RPA robot, by the computing system.

14. The computer-implemented method of claim 12 , wherein
the initial ML model is received from the conductor application via a proxy that is loosely coupled with the conductor application, the proxy configured to be decoupled from the conductor application and run independently therefrom, and
the proxy is configured to tunnel requests associated with the initial version of the ML model via an Application Programming Interface (API) gateway of a service tier.

15. The computer-implemented method of claim 12 , wherein when the ML model is a new version of an existing ML model, the method further comprises:
performing version control on the ML model, by the computing system, by creating and storing metadata regarding the new version of the ML model.

16. The computer-implemented method of claim 12 , further comprising:
uploading, validating, and deploying the ML model via a service tier, by the computing system.

17. The computer-implemented method of claim 16 , wherein the service tier is configured to provide internal utility services that perform asynchronous operations, state machine management, storage abstraction and access, or any combination thereof.

18. The computer-implemented method of claim 16 , wherein the service tier is configured to provide a model publish service that performs create, retrieve, update, and delete (CRUD) operations.

19. The computer-implemented method of claim 16 , wherein the service tier is configured to build images of the ML model with dependencies, build wrapper code around the ML model to create a container, push the container to a container registry, and deploy the container as an ML skill for consumption by the RPA robots.

20. The computer-implemented method of claim 12 , wherein the API is a Representative State Transfer (REST) API.

21. A system, comprising:
memory storing computer program instructions; and
at least one processor configured to execute the computer program instructions, wherein the computer program instructions are configured to cause the at least one processor to:
receive a machine learning (ML) model from a conductor application;
perform validation on the ML model; and
when validation of the ML model succeeds:
upload the ML model into storage,
deploy the ML model for use by robotic process automation (RPA) robots, and
publish the ML model by exposing the ML model as a service via a Representative State Transfer (REST) Application Programming Interface (API) that the RPA robots call during execution.

22. The system of claim 21 , wherein the computer program instructions are further configured to cause the at least one processor to:
receive a request from an RPA robot to execute the deployed ML model;
receive data to be used by the deployed ML model;
execute the deployed ML model using the received data; and
transmit results of the executed ML model to the RPA robot.

23. The system of claim 21 , wherein when the ML model is a new version of an existing ML model, the computer program instructions are further configured to cause the at least one processor to:
perform version control on the ML model by creating and storing metadata regarding the new version of the ML model.

24. The system of claim 21 , wherein the computer program instruction

Assignees

• Uipath Inc

Inventors

• Shrivastava Shashank
• Mcgonnell Anton

Classifications

• B25J9/163 Primary

  learning, adaptive, model based, rule based expert control · CPC title

• G06F8/60

  Software deployment · CPC title

• G06F8/71

  Version control (security arrangements therefor G06F21/57); Configuration management · CPC title

• G06F30/27

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G06N20/00 Primary

  Machine learning · CPC title