Granted Patent B2

US 12,254,387 · App. 18/478,964 · Granted Mar 18, 2025

Inventors: Kirk M. Lewis, II (New York, NY); Shahzad Din (New York, NY); Aarvi Thadeshwar (New York, NY); Christopher James Byrne (New York, NY)

Assignee: Citibank, N.A.

G06N20/00

Abstract

Claims (97)

1. A system for providing user interfaces using artificial intelligence, the system comprising:

one or more processors; and

one or more memories configured to store instructions that when executed by the one or more processors perform operations comprising:

receiving a request to display application data to a user of an application;

retrieving a user token associated with the user, wherein the user token comprises a plurality of user configurations generated based on user interface data received from a plurality of sources, and wherein the user token comprises the user interface data encoded into a vector space of a machine learning model;

generating a consolidated token based on the user token, wherein the consolidated token comprises a first subset of the user interface data and a second subset of a plurality of interface configurations;

inputting the consolidated token into the machine learning model to obtain a user interface token;

decoding the user interface token into a plurality of fields storing interface configuration settings;

retrieving a user interface configuration template of a plurality of user interface configuration templates, wherein the user interface configuration template is associated with the application;

matching the plurality of fields storing the interface configuration settings with a plurality of template fields to generate a unique user-application interface configuration; and

sending a command to the application to display the application data in a user interface, wherein the command causes the application to initialize the user interface with the unique user-application interface configuration.

2. A method for providing user interfaces using artificial intelligence, the method comprising:

receiving a request to display application data to a user of an application,

retrieving a user token associated with the user, wherein the user token comprises a plurality of user configurations generated based on user interface data received from a plurality of sources;

generating a consolidated token based on the user token, wherein the consolidated token comprises a first subset of the user interface data and a second subset of a plurality of interface configurations;

inputting the consolidated token into a machine learning model to obtain a user interface token;

decoding the user interface token into a plurality of configuration fields storing interface configuration settings;

retrieving a user interface configuration template of a plurality of user interface configuration templates, wherein the user interface configuration template is associated with the application;

matching the plurality of configuration fields storing the interface configuration settings with a plurality of template fields to generate a unique user-application interface configuration; and

sending a command to the application to display the application data in a user interface, wherein the command causes the application to initialize the user interface with the unique user-application interface configuration.

3. The method of claim 2 , further comprising:

receiving a plurality of interface configurations for the application;

generating a settings request for the application, wherein the settings request requests the plurality of interface configurations for the application;

receiving, in response to the settings request, the plurality of interface configurations for the application;

encoding the plurality of interface configurations into a vector within a vector space of the machine learning model; and

generating an application token from the vector.

4. The method of claim 3 , wherein encoding the plurality of interface configurations into the vector within the vector space of the machine learning model comprises:

retrieving a plurality of fields corresponding to available interface configurations;

matching each interface configuration of the plurality of interface configurations to a corresponding field of the plurality of fields; and

encoding the plurality of interface configurations into the vector according to the plurality of fields, wherein predetermined values are added to fields that are not modifiable.

5. The method of claim 4 , further comprising merging the first subset of the user interface data and the second subset of the interface configurations by:

iterating through each interface configuration to determine a type associated with each interface configuration;

determining, within the user token, a corresponding user configuration that matches each type;

adding user interface configurations that have matching user configurations to the consolidated token; and

encoding the consolidated token into the vector space of the machine learning model.

6. The method of claim 2 , further comprising:

receiving a first plurality of sets of user interface settings for a plurality of applications, wherein each set of the first plurality of sets has been configured by a corresponding user of a plurality of users;

determining a plurality of environmental conditions associated with each set of user interface settings, wherein the plurality of environmental conditions comprises time of day, user location, and temperature at the user location at the time of day;

retrieving a second plurality of sets of user characteristics, wherein each set of user characteristics is associated with the corresponding user of the plurality of users;

generating a training dataset comprising the first plurality of sets of the user interface settings, the plurality of environmental conditions, and the second plurality of sets of the user characteristics; and

inputting the training dataset into a training routine of a user token generation machine learning model to train the user token generation machine learning model to output user tokens, wherein each user token corresponds to a matching plurality of sets of the user interface settings.

7. The method of claim 6 , further comprising:

retrieving a user identifier of the user;

determining, based on the user identifier, that the user token for the user has not been generated yet;

determining a plurality of current environmental conditions and a set of user characteristics for the user;

inputting the plurality of current environmental conditions and the set of user characteristics into the user token generation machine learning model to obtain a corresponding plurality of user interface settings; and

generating the user token based on the corresponding plurality of user interface settings.

8. The method of claim 6 , further comprising:

determining that the user token for the user has not been generated yet;

retrieving a set of user characteristics for the user;

inputting the user characteristics into a similarity model to identify a similar user within a set of users having corresponding user tokens; and

generating the user token based on a matching user token associated with the similar user.

9. The method of claim 2 , further comprising:

receiving, from the application, a message indicating that the user updated one or more user interface settings, wherein the message comprises indications of the one or more user interface settings;

inputting the one or more user interface settings into a user token generation machine learning model to obtain an updated plurality of user interface settings; and

generating an updated user token based on the updated plurality of user interface settings.

10. The method of claim 2 , further comprising:

extracting, from the request an application token comprising a plurality of interface configurations for the application;

determining that the application token comprises a first hierarchy of settings, wherein the first hierarchy of settings comprises a first plurality of levels of detail for the plurality of interface configurations;

determining that the user token comprises a second hierarchy of settings, wherein the second hierarchy of settings comprises a second plurality of levels of detail for the plurality of user configurations, and wherein the second plurality of levels has more levels than the first plurality of levels; and

updating the user token to remove one or more levels to match the first plurality of levels of detail.

11. One or more non-transitory, computer-readable media for providing user interfaces using artificial intelligence, storing instructions thereon that cause one or more processors to perform operations comprising:

receiving a request to display application data to a user of an application;

retrieving a user token associated with the user, wherein the user token comprises a plurality of user configurations generated based on user interface data received from a plurality of sources;

generating a consolidated token based on the user token, wherein the consolidated token comprises a first subset of the user interface data and a second subset of a plurality of interface configurations;

inputting the consolidated token into a machine learning model to obtain a user interface token;

decoding the user interface token into a plurality of configuration fields storing interface configuration settings;

retrieving a user interface configuration template of a plurality of user interface configuration templates, wherein the user interface configuration template is associated with the application;

matching the plurality of configuration fields storing the interface configuration settings with a plurality of template fields to generate a unique user-application interface configuration; and

sending a command to the application to display the application data in a user interface, wherein the command causes the application to initialize the user interface with the unique user-application interface configuration.

12. The one or more non-transitory, computer-readable media of claim 11 , wherein the instructions further cause the one or more processors to perform operations comprising:

receiving a plurality of interface configurations for the application;

generating a settings request for the application, wherein the settings request requests the plurality of interface configurations for the application;

receiving, in response to the settings request, the plurality of interface configurations for the application;

encoding the plurality of interface configurations into a vector within a vector space of the machine learning model; and

generating an application token from the vector.

13. The one or more non-transitory, computer-readable media of claim 12 , wherein the instructions for encoding the plurality of interface configurations into the vector within the vector space of the machine learning model further cause the one or more processors to perform operations comprising:

retrieving a plurality of fields corresponding to available interface configurations;

matching each interface configuration of the plurality of interface configurations to a corresponding field of the plurality of fields; and

encoding the plurality of interface configurations into the vector according to the plurality of fields, wherein predetermined values are added to fields that are not modifiable.

14. The one or more non-transitory, computer-readable media of claim 12 , wherein the instructions for merging the application token and the user token into the consolidated token further cause the one or more processors to perform operations comprising:

iterating through each user interface configuration of the application token to determine a type associated with each user interface configuration;

determining, within the user token, a corresponding user configuration that matches each type;

adding user interface configurations that have matching user configurations to the consolidated token; and

encoding the consolidated token into the vector space of the machine learning model.

15. The one or more non-transitory, computer-readable media of claim 11 , further causing the one or more processors to perform operations comprising:

receiving a first plurality of sets of user interface settings for a plurality of applications, wherein each set of the first plurality of sets has been configured by a corresponding user of a plurality of users;

determining a plurality of environmental conditions associated with each set of user interface settings, wherein the plurality of environmental conditions comprises time of day, user location, and temperature at the user location at the time of day;

retrieving a second plurality of sets of user characteristics, wherein each set of user characteristics is associated with the corresponding user of the plurality of users;

generating a training dataset comprising the first plurality of sets of the user interface settings, the plurality of environmental conditions, and the second plurality of sets of the user characteristics; and

inputting the training dataset into a training routine of a user token generation machine learning model to train the user token generation machine learning model to output user tokens, wherein each user token corresponds to a matching plurality of sets of user interface settings.

16. The one or more non-transitory, computer-readable media of claim 15 , wherein the instructions further cause the one or more processors to perform operations comprising:

retrieving a user identifier of the user;

determining, based on the user identifier, that the user token for the user has not been generated yet;

determining a plurality of current environmental conditions and a set of user characteristics for the user;

inputting the plurality of current environmental conditions and the set of user characteristics into the user token generation machine learning model to obtain a corresponding plurality of user interface settings; and

generating the user token based on the corresponding plurality of user interface settings.

Assignments (2)

Continuity (2)

References Cited (15)