Granted Patent B1

US 12,314,321 · App. 18/658,081 · Granted May 27, 2025

Inventors: Ahamed Jalaldeen Shahul Hamid (Bangalore, IN); Mukundan Sundararajan (Bangalore, IN); Deepak Malik (Gurgaon, IN); Hemant Singh (Gurgaon, IN)

Assignee: International Business Machines Corporation

G06F16/9035

Abstract

Claims (71)

1. A computer-implemented method for automatically recommending appropriate foundation models, the computer-implemented method comprising:

generating, by a computer, using a foundation model advisor, a list of recommended foundation models, along with a corresponding ranking for each respective foundation model in the list based on a corresponding attribute matching score and a justification for the corresponding ranking of each respective foundation model;

determining a subset of foundation models based on the justification to form the list of recommended foundation models;

in response to the determining the subset of foundation models,

sending, by the computer, using the foundation model advisor, the list of the recommended foundation models, along with the corresponding ranking for each respective foundation model in the list and the justification for the corresponding ranking of each respective foundation model to a user;

receiving, by the computer, using the foundation model advisor, feedback from the user regarding accuracy of a user-selected foundation model from the list of the recommended foundation models; and

adjusting, by the computer, using a generative AI manager, the user-selected foundation model based on the feedback received from the user regarding the accuracy of the user-selected foundation model from the list of the recommended foundation models.

2. The computer-implemented method of claim 1 , further comprising:

deploying, by the computer, using the generative AI manager, the user-selected foundation model on a client device of the user to develop a generative AI solution in response to adjusting the user-selected foundation model based on the feedback.

3. The computer-implemented method of claim 1 , further comprising:

receiving, by the computer, using the foundation model advisor, project requirements corresponding to a generative AI solution from the user; and

performing, by the computer, using the foundation model advisor, a comparison of the project requirements received from the user with attributes of each respective foundation model of a plurality of foundation models contained in a foundation model catalog.

4. The computer-implemented method of claim 3 , further comprising:

generating, by the computer, using the foundation model advisor, an attribute matching score for each respective foundation model of the plurality of foundation models contained in the foundation model catalog based on the comparison of the project requirements received from the user with the attributes of each respective foundation model; and

selecting, by the computer, using the foundation model advisor, the recommended foundation models having the corresponding attribute matching score greater than a predefined minimum attribute matching score threshold level.

5. The computer-implemented method of claim 1 , further comprising:

identifying, by the computer, using robotic process automation, a plurality of foundation model hubs, each foundation model hub contains a plurality of foundation models along with attributes of corresponding foundation models;

retrieving, by the computer, using the robotic process automation, a list of foundation models from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model in response to identifying the plurality of foundation model hubs; and

generating, by the computer, using the generative AI manager of the computer, a foundation model catalog based on the list of foundation models retrieved from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model.

6. The computer-implemented method of claim 5 , further comprising:

training, by the computer, using the generative AI manager of the computer, a general-purpose foundation model of the computer using the foundation model catalog.

7. The computer-implemented method of claim 6 , further comprising:

implementing, by the computer, using the generative AI manager of the computer, the foundation model advisor of the computer based on the general-purpose foundation model that was trained with the foundation model catalog; and

adding, by the computer, using the generative AI manager of the computer, the general-purpose foundation model to the foundation model advisor of the computer.

8. The computer-implemented method of claim 6 , further comprising:

utilizing, by the computer, the foundation model advisor to generate the list of the recommended foundation models in accordance with received project requirements, along with a ranking corresponding to each particular foundation model in the list, and the justification in natural language for the ranking of each particular foundation model.

9. A computer system for automatically recommending appropriate foundation models, the computer system comprising:

a communication fabric;

a set of computer-readable storage media connected to the communication fabric, wherein the set of computer-readable storage media collectively stores program instructions; and

a set of processors connected to the communication fabric, wherein the set of processors executes the program instructions to:

generate, using a foundation model advisor, a list of recommended foundation models, along with a corresponding ranking for each respective foundation model in the list based on a corresponding attribute matching score and a justification for the corresponding ranking of each respective foundation model;

determining a subset of foundation models based on the justification to form the list of recommended foundation models;

in response to the determining the subset of foundation models,

send, using the foundation model advisor, the list of the recommended foundation models, along with the corresponding ranking for each respective foundation model in the list and the justification for the corresponding ranking of each respective foundation model to a user;

receive, using the foundation model advisor, feedback from the user regarding accuracy of a user-selected foundation model from the list of the recommended foundation models; and

adjust, using a generative AI manager, the user-selected foundation model based on the feedback received from the user regarding the accuracy of the user-selected foundation model from the list of the recommended foundation models.

10. The computer system of claim 9 , wherein the set of processors further executes the program instructions to:

deploy, using the generative AI manager, the user-selected foundation model on a client device of the user to develop a generative AI solution in response to adjusting the user-selected foundation model based on the feedback.

11. The computer system of claim 9 , wherein the set of processors further executes the program instructions to:

receive, using the foundation model advisor, project requirements corresponding to a generative AI solution from the user; and

perform, using the foundation model advisor, a comparison of the project requirements received from the user with attributes of each respective foundation model of a plurality of foundation models contained in a foundation model catalog.

12. The computer system of claim 11 , wherein the set of processors further executes the program instructions to:

generate, using the foundation model advisor, an attribute matching score for each respective foundation model of the plurality of foundation models contained in the foundation model catalog based on the comparison of the project requirements from the user with the attributes of each respective foundation model; and

select, using the foundation model advisor, the recommended foundation models having the corresponding attribute matching score greater than a predefined minimum attribute matching score threshold level.

13. The computer system of claim 9 , wherein the set of processors further executes the program instructions to:

identify, using robotic process automation, a plurality of foundation model hubs, each foundation model hub contains a plurality of foundation models along with attributes of corresponding foundation models;

retrieve, using the robotic process automation, a list of foundation models from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model in response to identifying the plurality of foundation model hubs; and

generate, using the generative AI manager of the computer system, a foundation model catalog based on the list of foundation models retrieved from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model.

14. The computer system of claim 13 , wherein the set of processors further executes the program instructions to:

train, using the generative AI manager of the computer system, a general-purpose foundation model of the computer system using the foundation model catalog.

15. A computer program product for automatically recommending appropriate foundation models, the computer program product comprising a set of computer-readable storage media having program instructions collectively stored therein, the program instructions executable by a computer to cause the computer to:

generate, using a foundation model advisor, a list of recommended foundation models, along with a corresponding ranking for each respective foundation model in the list based on a corresponding attribute matching score and a justification for the corresponding ranking of each respective foundation model;

determining a subset of foundation models based on the justification to form the list of recommended foundation models;

in response to the determining the subset of foundation models,

send, using the foundation model advisor, the list of the recommended foundation models, along with the corresponding ranking for each respective foundation model in the list and the justification for the corresponding ranking of each respective foundation model to a user;

receive, using the foundation model advisor, feedback from the user regarding accuracy of a user-selected foundation model from the list of the recommended foundation models; and

adjust, using a generative AI manager, the user-selected foundation model based on the feedback received from the user regarding the accuracy of the user-selected foundation model from the list of the recommended foundation models.

16. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

deploy, using the generative AI manager, the user-selected foundation model on a client device of the user to develop a generative AI solution in response to adjusting the user-selected foundation model based on the feedback.

17. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

receive, using the foundation model advisor, project requirements corresponding to a generative AI solution from the user; and

perform, using the foundation model advisor, a comparison of the project requirements received from the user with attributes of each respective foundation model of a plurality of foundation models contained in a foundation model catalog.

18. The computer program product of claim 17 , wherein the program instructions further cause the computer to:

generate, using the foundation model advisor, an attribute matching score for each respective foundation model of the plurality of foundation models contained in the foundation model catalog based on the comparison of the project requirements received from the user with the attributes of each respective foundation model; and

select, using the foundation model advisor, the recommended foundation models having the corresponding attribute matching score greater than a predefined minimum attribute matching score threshold level.

19. The computer program product of claim 15 , wherein the program instructions further cause the computer to:

identify, using robotic process automation, a plurality of foundation model hubs, each foundation model hub contains a plurality of foundation models along with attributes of corresponding foundation models;

retrieve, using the robotic process automation, a list of foundation models from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model in response to identifying the plurality of foundation model hubs; and

generate, using the generative AI manager of the computer, a foundation model catalog based on the list of foundation models retrieved from each respective foundation model hub of the plurality of foundation model hubs along with the attributes of each respective foundation model.

20. The computer program product of claim 19 , wherein the program instructions further cause the computer to:

train, using the generative AI manager of the computer, a general-purpose foundation model of the computer using the foundation model catalog.

Assignments (1)

References Cited (25)