Welcome To The Gen AI Multi Agentic IT support

Despite considerable progress in conversational AI and IT support automation, the review of existing literature and implemented systems highlights several persistent and interconnected research gaps. These gaps span four core domains—multi-agent systems, knowledge base management, form automation, and emotionally intelligent 3D avatars—underscoring the need for a more adaptive, context-aware, and human-centered IT support platform.

1. Multi-Agent System with MCP server

* Existing IT support systems rely on single-agent or static automation, lacking multi-agent coordination for task specialization and efficient distribution of complex support requests.

* Limited research exists on integrating multi-agent frameworks with modern LLMs for real-time inference, prioritization, and autonomous decision-making, which restricts scalability and adaptability.

2. Knowledge-base powered by episodic memory

* Current knowledge bases are static and manually updated, often becoming outdated and leading to longer resolution times in dynamic IT environments.

* LLM-driven systems frequently generate irrelevant or hallucinated responses due to weak memory management, limited token optimization, and lack of adaptive feedback loops.

3. Form Automation System

* Automated form-filling tools remain rigid, relying on predefined slots that cannot handle unstructured IT-specific queries, leading to user fatigue and incomplete tickets.

* Sinhala voice navigation system tailored to the needs of blind students.

* Existing solutions predominantly cater to English-speaking users, rendering them ineffective for Sinhala-speaking individuals who require navigation assistance in their native language.

4. 3D Avatar with Emotional Intelligence

* Current IT chatbots lack emotional intelligence, failing to integrate sentiment analysis or speech emotion recognition, which results in mechanical and unsympathetic responses.

* Few systems combine multimodal emotion sensing with dynamic 3D avatars, missing opportunities to enhance trust, engagement, and reassurance during high-stress IT scenarios.

The main research problem here is the difficulty of accessing printed materials with braille technology for visually impaired individuals.

There is no proper mobile application for sinhala book reading.

Obtaining educational materials and resources that meet the language needs of visually impaired Sinhala-speaking individuals is a significant challenge. The lack of a comprehensive educational access program designed specifically for this group of people poses major problems. Their educational opportunities are limited due to this deficiency, which in turn hinders their socioeconomic prospects.

Access to Sinhala-language educational resources is limited for speakers of Sinhala, as most assistive technology currently in use is designed for English-language materials. Additionally, translating printed Sinhala text into natural-sounding speech is challenging due to the intricate script and subtle syntax of the Sinhala language, which poses difficulties for text-to-speech (TTS) and optical character recognition (OCR) systems. Moreover, people with visual impairments face difficulties accessing visual content that is frequently included in educational resources, as a crucial component of picture interpretation is still absent.

Our research aims to create an inclusive educational access system that enables efficient access to educational content for visually impaired Sinhala speakers.

Main Objective

Our research goals are aimed at providing a comprehensive solution to the issue previously mentioned. Firstly, we aim to improve Memory and agents, For working with orchestrators to provide best results. This will enable these technologies to work together more effectively, resulting in a more efficient and accurate system.

It is important to make the knowledge base more dynamic and adaptable. This will involve developing methods for automatically updating and expanding the knowledge base, as well as improving the system's ability to learn from user interactions and feedback.

We aim to continuously improve our solution based on customer feedback and experiences. To achieve this, we plan to conduct comprehensive testing and collect user feedback. Our main objective is to remove the existing systems and make them automated so the support staff can work more efficient and would be able to come up with faster solutions for the employees' problems.

Sub Objectives

Integrate with MCP server for seamless data exchange and processing.

Connect all the other components as agents to multi-agentic system to work as a single system.

Implement feedback loops to continuously learn from user interactions.

Enhance the system's ability to handle diverse data sources and formats.

Develop a robust monitoring and evaluation framework to assess system performance.

The methodology for the IT support platform involves a mixed-methods approach. Firstly, we will conduct qualitative research by interviewing end users and employees, as well as holding focus groups, to gain insights into their needs, preferences, and challenges when using the platform. Secondly, we will carry out a quantitative phase by surveying a larger sample of users to gather data on usage patterns and evaluate the platform’s impact on productivity and issue resolution.

Additionally, we will conduct usability testing to assess the user-friendliness of the current platform. We will also review existing IT support platforms and tools to evaluate their features, performance, and service capabilities.

The combined data will be analyzed to identify areas for improvement, opportunities for customization, and enhanced functionality, which will guide the development of a more effective and user-centered IT support platform.

Sentence Transformers

Python 3

Hugging face

ReactJs

RASA frameworks

Postman API

VS Code

Docker

GitLab

MCP