Aggregation of a plurality of artificial intelligence knowledge models

1 . A method for aggregating knowledge from a plurality specialized knowledge models (KMs), the method comprising:

receiving a query from a user;

analyzing, using a query analyzer machine language model (MLM), the query to determine that the query is directed to two or more different subject matter areas;

selecting, from a KM registry comprising information about a plurality of specialized KMs each having a distinct subject matter expertise, the plurality of specialized KMs each having respective expertise in one of the two or more different subject matter areas of the query;

generating, using a query generator, a first sub-query directed to the first subject matter of the query and a second sub-query directed to the second subject matter of the query;

transmitting the first sub-query for processing by the selected first specialized KM having expertise in the subject matter of the first sub-query;

obtaining a first response from the first specialized KM based on the first sub-query;

transmitting the second sub-query and the first response from the first specialized KM for processing by the selected second specialized KM having expertise in the subject matter of the second sub-query; and

generating, for display on a user interface (UI), an aggregated response from the second specialized KM, wherein the aggregated response combines knowledge from both the first specialized KM and the second specialized KM.

2 . The method of claim 1 , wherein the query generator comprises a natural language processing MLM, and first sub-query and a second sub-query comprise two different natural language queries.

3 . The method of claim 2 , wherein the query generator comprises a MLM prepared for processing physical data models based at least in part on transmitting data in iterations or in a sequence.

4 . The method of claim 1 , wherein the query generator comprises an embeddings generator model, and the first sub-query and the second sub-query comprises two different embedding vectors.

5 . The method of claim 1 , wherein the first response and aggregated response include one of natural language or an embeddings vector.

6 . The method of claim 1 , further comprising:

generating, for display on the UI, each response from the first specialized KM or the second specialized KM.

7 . The method of claim 1 , wherein the first specialized KM and the second specialized KM correspond to at least one of a Retrieval-Augmented Generation (RAG), Low-Rank Adaption (LORA), or Graph Neural Networks (GNN).

8 . The method of claim 1 , further comprising:

transmitting a request to an application programming interface (API) for the first specialized KM with the query, wherein the API processes the query by routing the query to the first specialized KM for processing the query and generating the first response; and

transmitting a request to an API for the second specialized KM with the query and the first response, wherein the API processes the query by routing the query and the first response to the second specialized KM for processing the query and generating the aggregated response.

9 . The method of claim 1 , further comprising:

training the query analyzer MLM to determine if the query is directed to two or more different subject matter areas based on using a training dataset comprising at least queries including examples of single-subject queries and multi-subject queries.

10 . The method of claim 1 , analyzing the query to determine if the query is directed to two or more different subject matter areas by:

performing a tokenization process by breaking down the query into tokens for the query analyzer MLM model to process; and

using n-grams to capture context and relationships between words.

11 . The method of claim 1 , further comprising:

training the first specialized KM having expertise in the subject matter of the first sub-query based on using a training dataset comprising domain-specific data in the subject matter of the first sub-query to master at least one of language, concepts, or reasoning within a particular field; and

training the second specialized KM having expertise in the subject matter of the second sub-query based on using a training dataset comprising domain-specific data in the subject matter of the second sub-query to master at least one of language, concepts, or reasoning within a particular field.

12 . The method of claim 1 , wherein the query generator is configured to:

process the query and identify distinct subject matters using the query generator,

classify the distinct subject matters by using a Latent Dirichlet Allocation (LDA) or pre-trained topic classifier to determine the subject of different segments of the query, and

generate sub-queries for each subject matter.

13 . The method of claim 1 , wherein the information for specialized KM comprises at least one of:

a model name, subject matter expertise that the specialized KM is trained to handle, a version or iteration of the specialized KM, base model architecture, model type, fine-tuning dataset used to fine-tune the specialized KM for its specific subject matter expertise, a number of parameters, accuracy/precision/recall/F1 score, benchmarks, inference speed, resource requirements, training data sources, training time, hardware used, optimization techniques, or API detail, deployment requirements, update date, or license information.

14 . The method of claim 1 , wherein the specialized KM is an object comprising a MLM with application programming interfaces (API) for questions-answers and associated annotation from the KM registry.

15 . The method of claim 1 , wherein transmitting the second sub-query and the first response from the first specialized KM further comprises:

in response to obtaining a first prompt from the user, adding the query and the obtained first response from the first specialized KM to the second sub-query; and

in response to obtaining a second prompt from the user, adding a prompt for answering the second sub-query in context with the obtained first response from the first specialized KM to the second sub-query.

16 . A system for aggregating knowledge from a plurality specialized knowledge models, the system comprising:

at least one memory; and

at least one hardware processor coupled with the at least one memory and configured, individually or in combination, to:

receive a query from a user;

analyze, using a query analyzer machine language model (MLM), the query to determine that the query is directed to two or more different subject matter areas;

select, from a KM registry comprising information about a plurality of specialized KMs each having a distinct subject matter expertise, the plurality of specialized KMs each having respective expertise in one of the two or more different subject matter areas of the query;

generate, using a query generator, a first sub-query directed to the first subject matter of the query and a second sub-query directed to the second subject matter of the query;

transmit the first sub-query for processing by the selected first specialized KM having expertise in the subject matter of the first sub-query;

obtain a first response from the first specialized KM based on the first sub-query;

transmit the second sub-query and the first response from the first specialized KM for processing by the selected second specialized KM having expertise in the subject matter of the second sub-query; and

generate, for display on a user interface (UI), an aggregated response from the second specialized KM, wherein the aggregated response combines knowledge from both the first specialized KM and the second specialized KM.

17 . The system of claim 16 , wherein the query generator comprises a natural language processing MLM, and first sub-query and a second sub-query comprise two different natural language queries.

18 . The system of claim 16 , wherein the query generator includes an embeddings generator model, and first sub-query and a second sub-query include two different embedding vectors.

19 . The system of claim 16 , wherein the first response and aggregated response include one of natural language or an embeddings vector.

20 . A non-transitory computer readable medium storing thereon computer executable instructions for aggregating knowledge from a plurality specialized knowledge models, including instructions for:

receiving a query from a user;

analyzing, using a query analyzer machine language model (MLM), the query to determine that the query is directed to two or more different subject matter areas;

selecting, from a KM registry comprising information about a plurality of specialized KMs each having a distinct subject matter expertise, the plurality of specialized KMs each having respective expertise in one of the two or more different subject matter areas of the query;

generating, using a query generator, a first sub-query directed to the first subject matter of the query and a second sub-query directed to the second subject matter of the query;

transmitting the first sub-query for processing by the selected first specialized KM having expertise in the subject matter of the first sub-query;

obtaining a first response from the first specialized KM based on the first sub-query;

transmitting the second sub-query and the first response from the first specialized KM for processing by the selected second specialized KM having expertise in the subject matter of the second sub-query; and

generating, for display on a user interface (UI), an aggregated response from the second specialized KM, wherein the aggregated response combines knowledge from both the first specialized KM and the second specialized KM.