Description:FIELD OF INVENTION
The present invention generally relates to the field of computational problem-solving. More specifically, it pertains to a system and method of orchestrating dynamic persona-based agents for complex problem solving for dynamically coordinating multiple artificial intelligence agents, leveraging large language models (LLMs) and small language models (SLMs), to address complex tasks that require contextual understanding, adaptability, and multi-agent collaboration.

BACKGROUND:
Advancements in artificial intelligence have given rise to powerful language models like LLMs, which are capable of generating human-like text, and SLMs, which are fine-tuned for specific, narrowly defined tasks. These models have found applications in diverse fields, from legal research to medical diagnostics and creative writing. However, as the scope of problems grows more complex, requiring a mix of broad contextual knowledge and niche expertise, existing systems face significant limitations.
Traditional AI-based systems typically rely on a single model or a rigid ensemble of models to handle tasks. While this approach may work for straightforward problems, it becomes inadequate for complex scenarios requiring specialized knowledge and collaboration among multiple experts. For instance, solving a legal case involving technical patents may require expertise in law, technical writing, and patent landscaping, all of which demand different types of knowledge and reasoning. Current systems lack the flexibility to dynamically allocate the right resources and coordinate their efforts efficiently.
Moreover, the lack of adaptability and contextual understanding further exacerbates these limitations. Static systems often fail to adjust their approach based on historical insights, evolving problem requirements, or the need for real-time collaboration among multiple specialized agents. This leads to inefficiencies, suboptimal outcomes, and limited scalability when dealing with multi-domain or interdisciplinary challenges.
Recognizing these challenges, the present invention introduces a novel system that dynamically orchestrates specialized persona-based agents. By leveraging the strengths of both LLMs and SLMs, contextual data, and historical insights, the system generates and coordinates task-specific personas that collaborate to solve complex problems. This approach ensures optimal utilization of resources, improved contextual understanding, and seamless multi-agent collaboration, thereby addressing the shortcomings of traditional AI-based systems.
PRIOR ART
WO2020149172A1 relates to the construction of an agent capable of responding to complex tasks by utilizing a value function. The approach uses a weighted sum of component tasks, wherein each component agent is tasked with solving a portion of the problem, and their actions are coordinated based on the value functions learned for each task. While effective in task delegation and action formulation, this approach does not specifically address the need for dynamic coordination and context-based adaptation in real-time problem-solving, particularly when specialized agent personas are required to handle tasks that demand diverse expertise and collaboration.
CN108009012A describes a system and method for facilitating collaboration among multiple users or groups to solve problems. It focuses on the specification of problem, conclusion, and knowledge structures, and aims to enable collaborative problem solving. However, it does not incorporate the use of specialized persona-based agents, nor does it address the dynamic orchestration of agents to handle tasks with contextual understanding or real-time data integration, which are essential for solving more complex, domain-specific problems.
In summary, the present invention provides a significant advancement over the prior art by addressing the specific challenge of dynamically orchestrating specialized, persona-based agents for real-time, complex problem-solving.

DEFINITIONS:
The expression "system" used hereinafter in this specification refers to an ecosystem comprising, but is not limited to, a Gateway Interface, Large Language Models (LLMs), Fine-Tuned Small Language Models (SLMs), a Probing Layer, a Contextual Data Repository, a Persona Selector, and a Collaboration Module, each working in a coordinated manner to enable the execution of inferencing tasks and facilitate iterative interactions among personas for task completion. The system further encompasses computing systems, including but not limited to mobile devices, laptops, desktops, and servers.
The expression "Gateway Interface" refers to a component acting as the primary entry point for inferencing requests within the system. This interface is responsible for managing and directing incoming requests to the appropriate Large Language Models (LLMs) for the identification of intents and attributes within the requests. It ensures the correct redirection and management of tasks to facilitate smooth and accurate inferencing processes within the system.
The expression "Large Language Models" refers to an advanced computational model employed by the system to perform generic inferencing tasks. The LLMs serve as the core processing engines to identify the intents and attributes of incoming requests. In addition to intent identification, the LLMs also provide initial suggestions for predefined instructions that could potentially be used to execute the tasks, based on the analysis of the request content.
The expression "Fine-Tuned Small Language Models" refers to a specialized subset of language models within the system, designed to address specific tasks by leveraging the intents and attributes identified by the Large Language Models. These SLMs are specifically trained and fine-tuned to solve particular types of tasks, thereby providing a targeted approach to inferencing that enhances the overall efficiency and effectiveness of task execution within the system.
The expression "Probing Layer" refers to a specialized analytical component within the system responsible for evaluating combinations of identified intents and attributes using the Large predefined instructions that should be employed to execute the tasks effectively. The probing layer plays a critical Language Models. This layer is tasked with analyzing these combinations to determine the most appropriate ethical role in refining the inferencing process by ensuring that the right predefined instructions are selected based on intent and attribute analysis.
The expression "Contextual Data Repository" refers to a centralized storage component within the system that aggregates a wide variety of data from diverse sources, including code repositories, project management tools, and relevant documentation. This repository serves as a historical data store, providing valuable insights and context that can be used to validate and refine persona selections. The repository enhances the accuracy and relevance of inferencing by making historical data available for analysis.
The expression "Persona Selector" refers to a critical component within the system designed to identify and select the personas typically involved in specific tasks based on the analysis conducted by the Large Language Models. The Persona Selector is responsible for using historical data, stored within the Contextual Data Repository, to validate and refine its persona selection process. This ensures that the right individuals or roles are chosen to address and complete the tasks efficiently, based on their expertise and involvement in similar activities.
The expression "Collaboration Module" refers to a component within the system that facilitates coordinated interaction among the selected personas for the execution of tasks. This module enables personas to collaborate iteratively, ensuring that tasks are executed in a systematic and organized manner. The Collaboration Module ensures seamless communication, task delegation, and progress tracking, thereby fostering a collaborative environment for the efficient completion of tasks across multiple personas within the system.

OBJECTS OF THE INVENTION:
The primary objective of this invention is to provide a system and method of orchestrating dynamic persona-based agents for complex problem solving.
Yet another object of the present invention is to utilize large language models (LLMs) for generic inferencing and intent identification, ensuring that the system can understand and interpret a wide range of problem scenarios.
Yet another object of the present invention is to employ fine-tuned small language models (SLMs) for executing specific tasks, enabling precise and specialized handling of domain-specific challenges.
Yet another object of the present invention is to integrate contextual data from diverse sources, such as code repositories, project management systems, and documentation, to ensure that the agents' actions are informed by the most relevant and up-to-date information.
Yet another object of the present invention is to employ a gateway interface that efficiently manages inferencing requests and routes them to the appropriate models, optimizing the task-solving process.
Yet another object of the present invention is to use historical data to validate and refine the selection of personas involved in task-solving, improving the system's effectiveness and adaptability over time.
Further the object of the present invention is to facilitate iterative collaboration among selected personas, enabling them to collaboratively address and solve tasks until the entire complex problem is resolved.

SUMMARY
Before the present invention is described, it is to be understood that the present invention is not limited to specific methodologies and materials described, as these may vary as per the person skilled in the art. It is also to be understood that the terminology used in the description is for the purpose of describing the particular embodiments only and is not intended to limit the scope of the present invention.
The present invention relates to a dynamic, persona-based agent orchestration system designed to solve complex problems through a collaborative, iterative approach involving both human and computer- implemented agents. The system begins by receiving an inferencing request, which includes a problem description and contextual data, and proceeds through several steps to identify, analyze, and resolve the problem. Initially, a Large Language Model (LLM) parses the problem description to identify its core intent and key attributes. The system then uses the LLM to suggest suitable predefined instructions, predefined instructions, or techniques to address the problem. Furthermore, the system identifies and refines personas—human roles or expertise—necessary to solve the problem. These personas are instantiated as virtual agents and are assigned specific subtasks based on their expertise and the identified instructions. Fine-Tuned Small Language Models (SLMs) are used to empower these agents with specialized capabilities to efficiently execute their tasks. The system facilitates iterative execution, where personas collaborate dynamically, sharing insights and adjusting their actions in response to progress and emerging challenges, until the problem is fully resolved. This invention effectively orchestrates a seamless collaboration between human expertise and AI, enhancing the efficiency and adaptability of problem-solving in complex environments.

BRIEF DESCRIPTION OF DRAWINGS
FIG. 1 illustrates an overview of the system of the present invention.
FIG. 2 illustrates a stepwise method for collaborative problem-solving, employed by the system of the present invention.

DETAILED DESCRIPTION OF INVENTION:
Before the present invention is described, it is to be understood that this invention is not limited to methodologies described, as these may vary as per the person skilled in the art. It is also to be understood that the terminology used in the description is for the purpose of describing the particular embodiments only and is not intended to limit the scope of the present invention. Throughout this specification, the word “comprise”, or variations such as “comprises” or “comprising”, will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps. The use of the expression “at least” or “at least one” suggests the use of one or more elements or ingredients or quantities, as the use may be in the embodiment of the invention to achieve one or more of the desired objects or results. Various embodiments of the present invention are described below. It is, however, noted that the present invention is not limited to these embodiments, but rather the intention is that modifications those are apparent are also included.
The present invention is directed to a system and method of orchestrating dynamic persona-based agents for complex problem solving; as illustrated in FIG. 1. wherein the system (1) comprises an input unit (2), a processing unit (3) and an output device (4) such that the processing unit (3) employs a well-defined workflow as illustrated in FIG. 2.
In accordance with an embodiment of the present invention , the method includes the steps as follows:
1. Inferencing Request (100) - wherein, a system receives an inferencing request at a gateway interface. The inferencing request comprises a problem description, which may be provided in various formats such as natural language text, structured data, or a combination thereof. The problem description may include a high-level overview of the problem, specific objectives or goals, relevant constraints, and any available background information or context. The system may utilize various techniques to parse and interpret the problem description, such as natural language processing (NLP) predefined instructions, semantic analysis, and data extraction methods. The parsed problem description is then processed and stored in a suitable format for subsequent steps in the problem-solving process. This step effectively captures the user's problem statement and initiates the collaborative problem-solving workflow.
2. Intent and Attribute Identification (200) - Subsequent to receiving the inferencing request, the system proceeds to the intent and attribute identification step (200). In this step, the system redirects the received problem description to a Large Language Model (LLM). The LLM, trained on a vast corpus of text and code, is employed to identify the underlying intent of the problem. This involves analyzing the problem description to determine the core objective or goal that the user seeks to achieve. Concurrently, the LLM extracts key attributes or characteristics of the problem that are relevant to its resolution. These attributes may include factors such as data sources, constraints, dependencies, and other pertinent information. The LLM leverages its advanced natural language processing capabilities to understand the nuances of human language and extract the essential information from the problem description. The identified intent and extracted attributes serve as critical inputs for subsequent steps in the problem-solving process, guiding the selection of appropriate predefined instructions and personas. This step effectively transforms the raw problem description into a structured representation that can be effectively processed by the system.
3. Probing Layer and Predefined Instructions (300) step - Wherein the system proceeds to the Identifying Predefined Instructions step. In this step, the system leverages the identified intent and extracted attributes to query the Large Language Model (LLM). The LLM, based on its training data and understanding of the problem domain, suggests a set of predefined instructions or predefined instructions suitable for addressing the identified problem. These instructions may encompass a wide range of techniques, including data analysis methods, machine learning predefined instructions, optimization techniques, and problem-solving heuristics. The LLM analyzes the intent and attributes to determine which instructions or predefined instructions are most likely to yield a successful solution. The suggested instructions are then presented to the system for further consideration and execution. This step effectively leverages the LLM's knowledge and expertise to propose a set of potential solutions to the identified problem, thereby guiding the subsequent problem-solving process.
4. Persona Identification (400) step - In this step, the system leverages the Large Language Model (LLM) to identify the personas typically involved in tasks related to the identified intent and the suggested instructions. Personas represent human roles or expertise that are relevant to the problem-solving process. The LLM, drawing upon its knowledge base and understanding of human roles and responsibilities within various domains, identifies the personas that possess the necessary skills, experience, and knowledge to effectively contribute to the problem-solving process. The identified personas provide a framework for organizing and delegating tasks, enabling efficient and effective collaboration among human and AI components within the problem-solving process. This step effectively leverages the LLM's understanding of human roles and expertise to identify the most suitable personas for the given problem, thereby enhancing the efficiency and effectiveness of the collaborative problem-solving process.
5. Persona Collaboration (500) step - In this step, the system instantiates the identified personas as agents within the problem-solving environment. These instantiated personas represent virtual entities that can interact with each other and with the system to execute the problem-solving tasks. The system then allocates subtasks to each persona based on their identified expertise and the predefined instructions. This allocation process may involve analyzing the personas' skills, experience, and preferences, as well as the requirements and dependencies of the subtasks. The system may utilize various techniques to allocate subtasks, such as task decomposition, workload balancing, and constraint satisfaction. Furthermore, the system leverages Fine-Tuned Small Language Models (SLMs) to empower each persona with specialized capabilities for executing their assigned subtasks. These SLMs are trained on specific domains or tasks and provide personas with enhanced abilities in areas such as data analysis, information retrieval, and communication. By leveraging SLMs, the system enhances the efficiency and accuracy of persona-specific subtask execution.
6. Iterative Execution phase (600) - In this phase, the instantiated personas execute their assigned subtasks, leveraging the capabilities of the Fine-Tuned Small Language Models (SLMs). As each persona completes a subtask, the system dynamically assesses the progress made and adjusts the subsequent execution steps. This iterative process involves continuous communication and information sharing among the personas. Intermediate results and insights generated by each persona are shared with the other collaborating personas, enabling them to refine their own subtasks and adjust their approaches accordingly. The system may also dynamically re-allocate subtasks based on the evolving state of the problem, the progress made by each persona, and any unforeseen challenges encountered during execution. This iterative approach allows the problem-solving process to adapt and evolve in response to new information, unforeseen circumstances, and the dynamic nature of the problem itself. The system continues to iterate through this process, with personas dynamically adjusting their roles and responsibilities, until the problem is fully resolved and the desired outcome is achieved.

WORKING EXAMPLE:
The invention hereafter will be cited by way of examples only for a better and detailed understanding.
A project manager seeks to analyze bottlenecks in a recent project's delivery timeline.
Step 1: Request Parsing
The system receives the following inferencing request:
● P (Problem Description): "Analyze bottlenecks in project delivery."
● C (Contextual Data): Includes:
○ Delivery Timeline: Project A timeline from 2022-01 to 2022-06.
○ Historical Delays: Phases 2 and 3 were delayed by 20 days each.
Step 2: Intent and Attribute Identification
The system redirects the problem description (P) to the LLM. The LLM identifies:
● III (Intent): "Identify bottleneck phases in delivery timeline."
● AAA (Attributes): {Phases, Team Availability, Timeline}
Step 3: Algorithm Identification
The system passes III and AAA to the probing layer. The LLM suggests the Critical Path Method (CPM) as a suitable algorithm.
Step 4: Persona Identification
The LLM initially suggests personas PLLM = {Project Manager, Data Analyst}.
Refining PLLM with historical data from C, the system identifies PR (Refined Personas) = {Data Analyst, QA Engineer}.
Step 5: Persona Collaboration
The system instantiates Data Analyst and QA Engineer as agents. Subtasks are allocated:
● Data Analyst: Computes delays for each phase using the CPM formula:
○ Delay = Actual Completion Time - Planned Completion Time
● QA Engineer: Validates delay analysis using quality reports.
Step 6: Iterative Execution
● Data Analyst: Computes delays:
○ Phase 2 Delay = 40 - 20 = 20 days
○ Phase 3 Delay = 50 - 30 = 20 days
● QA Engineer: Validates delays using quality reports and confirms team availability issues during Phases 2 and 3.
Output: The system presents the analysis, identifying Phases 2 and 3 as significant bottlenecks due to team availability issues.
Key Refinements:
● Explicitly used the provided algorithm and example.
● Incorporated specific formulas and calculations.
● Enhanced clarity and flow by using a narrative style.
● Focused on the practical application of each step.
This refined answer provides a more concrete and illustrative demonstration of the workflow, making it more impactful and relevant within the context of the provided algorithm and example.

While considerable emphasis has been placed herein on the specific elements of the preferred embodiment, it will be appreciated that many alterations can be made and that many modifications can be made in preferred embodiment without departing from the principles of the invention. These and other changes in the preferred embodiments of the invention will be apparent to those skilled in the art from the disclosure herein, whereby it is to be distinctly understood that the foregoing descriptive matter is to be interpreted merely as illustrative of the invention and not as a limitation.
, Claims:CLAIMS:
We claim,
1. A system and method for system and method of orchestrating dynamic persona-based agents for complex problem solving, wherein the system (1) comprises an input device (2), a processing unit (3) and an output device (4);
characterized in that:
the processing unit (3) of the system (1) employs a method comprising the steps of;
a. inferencing request (100) - receiving an inferencing request at a gateway interface, wherein the request includes a problem description and contextual data;
b. identifying intent and key attributes (200) associated with the problem using a large language model (LLM);
c. probing layer and predefined instructions (300) that includes probing a layer with the identified intent and attributes to suggest suitable predefined instructions;
d. identifying persona and refinement (400) that includes identifying personas typically involved in tasks related to the intent and refining the list of personas using historical data from the contextual data;
e. instantiating persona and allocating subtask (500) including instantiating the refined personas as agents and allocating subtasks to them, using fine-tuned small language models (SLMs) for persona-specific subtasks;
f. iteratively executing tasks (600) that includes executing the subtasks, with each persona forwarding results to collaborating personas and dynamically adjusting their roles and responsibilities based on intermediate results and the evolving problem context, until a solution to the problem is fully constructed.

2. The system as claimed in claim 1, wherein the contextual data comprises at least one of code repositories, documentation, historical project data, and user profiles.

3. The system as claimed in claim 1, wherein the LLM is a transformer-based model, and the probing layer comprises a knowledge graph of predefined instructions and their associated intents and attributes.

4. The system as claimed in claim 1, wherein persona identification includes analyzing historical performance data of individuals on similar tasks, and SLMs are fine-tuned on specific domains, such as data analysis, natural language processing, and software development.

5. The system as claimed in claim 1, wherein iterative execution includes dynamically adjusting the allocation of resources to personas based on their performance and the evolving needs of the problem, and the system provides a user interface for monitoring the progress of the problem-solving process and interacting with the personas.

6. The system as claimed in claim 1, wherein the system generates reports summarizing the problem-solving process, including the contributions of each persona, the predefined instructions used, and the final solution.

7. The system as claimed in claim 1, wherein the system is applied to at least one of application development, medical research, and financial analysis.
Dated this 6th day of January, 2025.