SYSTEMS AND METHODS FOR ROUTING MACHINE-LEARNING PROMPTS IN A DISTRIBUTED NETWORKING ENVIRONMENT

1 . A system, comprising:

one or more processors coupled to non-transitory memory, the one or more processors configured to:

maintain, in one or more data structures, a plurality of wagering opportunities, each wagering opportunity corresponding to a respective sport domain of a plurality of sport domains;

receive, from a client device for a communication session, a prompt identifying a sports wagering request;

determine, based on the prompt, a sport domain from the plurality of sport domains and a classification of an intent corresponding to a first output type of a plurality of output types;

select a first language model of a plurality of language models based on the sport domain and the classification of the intent of the prompt, the first language model fine-tuned on a first sport domain and a first classification of an intent such that the first language model is associated with a first sport and a first intent type, the first language model selected responsive to the sport domain and the classification of the intent matching the first sport and the first intent type of the first language model;

select, from the plurality of wagering opportunities, based on i) the prompt, ii) the sport domain, and iii) the classification of the intent, a subset of wagering opportunities corresponding to the sport domain and the classification of the intent;

generate, based on i) the prompt, ii) the sport domain, iii) the classification of the intent, and iv) the first language model, an input context comprising the prompt and identifying the subset of wagering opportunities; and

generate an output message using the first language model and the input context, the output message comprising data.

2 . The system of claim 1 , wherein the one or more processors are to:

determine the classification of the intent using a machine-learning model.

3 . The system of claim 2 , wherein the machine-learning model comprises a second language model of the plurality of language models, the second language model different from the first language model.

4 . The system of claim 1 , wherein the one or more processors are to:

determine the classification of the intent based on a set of predetermined keywords.

5 . The system of claim 1 , wherein the one or more processors are to:

generate a data structure indicating the first language model is associated with the first intent type based on a plurality of historical prompts and a corresponding plurality of historical output messages generated by the first language model.

6 . The system of claim 1 , wherein the one or more processors are to:

maintain a plurality of adapters each respectively corresponding to the plurality of language models, a first adapter of the plurality of adapters corresponding to the first language model; and

apply the first adapter to a base language model to generate the first language model.

7 . The system of claim 6 , wherein the first adapter comprises a low-rank adaptation data structure or a quantized low-rank adaptation data structure.

8 . The system of claim 1 , wherein a second intent type of a second language model of the plurality of language models corresponds to information requests, and wherein the first intent type of the first language model corresponds to recommendation requests.

9 . The system of claim 1 , wherein the one or more processors are to:

receive a second prompt corresponding to a second classification of a second intent;

select a second language model of the plurality of language models based on the second classification of the second intent, the second language model associated with a second intent type, the second language model selected responsive to the second classification of the second intent matching the second intent type of the second language model; and

generate a second output message using the second prompt and the second language model.

10 . The system of claim 9 , wherein the one or more processors are to:

provide the output message to the client device for presentation in a graphical user interface in response to the prompt; and

provide the second output message to the client device for presentation in the graphical user interface in response to the second prompt.

11 . A method, comprising:

maintaining, by one or more processors coupled to non-transitory memory, in one or more data structures, a plurality of wagering opportunities, each wagering opportunity corresponding to a respective sport domain of a plurality of sport domains;

receiving, by the one or more processors, from a client device for a communication session, a prompt identifying a sports wagering request;

determining, by the one or more processors, based on the prompt, a sport domain from the plurality of sport domains and a classification of an intent corresponding to a first output type of a plurality of output types;

selecting, by the one or more processors, a first language model of a plurality of language models based on the sport domain and the classification of the intent of the prompt, the first language model fine-tuned on a first sport domain and a first classification of an intent such that the first language model is associated with a first sport and a first intent type, the first language model selected responsive to the sport domain and the classification of the intent matching the first sport and the first intent type of the first language model;

selecting, by the one or more processors, from the plurality of wagering opportunities, based on i) the prompt, ii) the sport domain, and iii) the classification of the intent, a subset of wagering opportunities corresponding to the sport domain and the classification of the intent;

generating, by the one or more processors, based on i) the prompt, ii) the sport domain, iii) the classification of the intent, and iv) the first language model, an input context comprising the prompt and identifying the subset of wagering opportunities; and

generating, by the one or more processors, an output message using the first language model and the input context, the output message comprising data.

12 . The method of claim 11 , further comprising:

determining, by the one or more processors, the classification of the intent using a machine-learning model.

13 . The method of claim 12 , wherein the machine-learning model comprises a second language model of the plurality of language models, the second language model different from the first language model.

14 . The method of claim 11 , further comprising:

determining, by the one or more processors, the classification of the intent based on a set of predetermined keywords.

15 . The method of claim 11 , further comprising:

generating, by the one or more processors, a data structure indicating the first language model is associated with the first intent type based on a plurality of historical prompts and a corresponding plurality of historical output messages generated by the first language model.

16 . The method of claim 11 , further comprising:

maintaining, by the one or more processors, a plurality of adapters each respectively corresponding to the plurality of language models, a first adapter of the plurality of adapters corresponding to the first language model; and

applying, by the one or more processors, the first adapter to a base language model to generate the first language model.

17 . The method of claim 16 , wherein the first adapter comprises a low-rank adaptation data structure or a quantized low-rank adaptation data structure.

18 . The method of claim 11 , wherein a second intent type of a second language model of the plurality of language models corresponds to information requests, and wherein the first intent type of the first language model corresponds to recommendation requests.

19 . The method of claim 11 , further comprising:

receiving, by the one or more processors, a second prompt corresponding to a second classification of a second intent;

selecting, by the one or more processors, a second language model of the plurality of language models based on the second classification of the second intent, the second language model associated with a second intent type, the second language model selected responsive to the second classification of the second intent matching the second intent type of the second language model; and

generating, by the one or more processors, a second output message using the second prompt and the second language model.

20 . The method of claim 19 , further comprising:

providing, by the one or more processors, the output message to the client device for presentation in a graphical user interface in response to the prompt; and

providing, by the one or more processors, the second output message to the client device for presentation in the graphical user interface in response to the second prompt.