Project: Adaptive Multi‑Agent Reasoning System for Microsoft Certification Preparation

Track

Reasoning Agents (Azure AI Foundry)

Project Name

Adaptive Multi‑Agent Reasoning System for Microsoft Certification Preparation

GitHub Username

@xenon1919

Repository URL

http://github.com/xenon1919/Adaptive-Multi-Agent-Reasoning-System-for-Microsoft-Certification-Preparation/

Project Description

Adaptive Multi-Agent Reasoning System for Microsoft Certification Preparation is a production-style AI orchestration framework designed to guide learners preparing for Microsoft AI certifications (AI-102, DP-100).

Certification candidates often struggle with syllabus overload, lack of structured roadmaps, and no objective readiness evaluation. This system solves that by coordinating multiple specialized AI agents in a structured reasoning pipeline.

The architecture implements a planner–executor pattern with critic validation and adaptive feedback loops. A user prompt is first structured into JSON by the Input Structuring Agent. The Learning Curator Agent maps goals to relevant Microsoft Learn modules. The Study Planner Agent generates a realistic, milestone-driven weekly plan. The Assessment Agent evaluates readiness through exam-style questions and scoring. A Critic Agent validates plan quality, coverage completeness, and assessment rigor. Finally, a Decision Engine dynamically regenerates plans if readiness or coverage thresholds are not met.

Key features include:

• Role-based agent specialization
• Adaptive regeneration logic
• Independent critic validation
• Structured JSON schema guards
• Evaluation harness with multiple test scenarios

The system demonstrates robust multi-agent collaboration and adaptive orchestration suitable for real-world certification preparation workflows.

Demo Video or Screenshots

Screenshots: https://github.com/xenon1919/Adaptive-Multi-Agent-Reasoning-System-for-Microsoft-Certification-Preparation/blob/main/Screenshot%202026-02-22%20191433.png
https://github.com/xenon1919/Adaptive-Multi-Agent-Reasoning-System-for-Microsoft-Certification-Preparation/blob/main/Screenshot%202026-02-22%20191421.png

Primary Programming Language

Python

Key Technologies Used

• Python
• Azure OpenAI Service
• Multi-Agent Orchestration Architecture
• Structured JSON Validation
• Adaptive Workflow Engine
• Evaluation Harness Framework

Submission Type

Individual

Team Members

No response

Submission Requirements

• My project meets the track-specific challenge requirements
• My repository includes a comprehensive README.md with setup instructions
• My code does not contain hardcoded API keys or secrets
• I have included demo materials (video or screenshots)
• My project is my own work with proper attribution for any third-party code
• I agree to the Code of Conduct
• I have read and agree to the Disclaimer
• My submission does NOT contain any confidential, proprietary, or sensitive information
• I confirm I have the rights to submit this content and grant the necessary licenses

Quick Setup Summary

1. Clone the repository:
   git clone https://github.com/xenon1919/Adaptive-Multi-Agent-Reasoning-System-for-Microsoft-Certification-Preparation.git

2. Navigate to project directory

3. Install dependencies:
   pip install -r requirements.txt

4. Configure environment variables:

   • Add Azure OpenAI credentials in .env (see .env.example)

5. Run main workflow:
   python main.py

6. Run evaluation harness:
   python run_evaluation.py

Technical Highlights

• Designed a true planner–executor–critic architecture rather than single-prompt chaining.
• Implemented adaptive regeneration logic based on readiness and coverage thresholds.
• Built a structured JSON validation layer to reduce hallucination and enforce schema integrity.
• Developed an evaluation harness with multiple real-world certification scenarios.
• Integrated a decision engine to dynamically adjust learning plans based on assessment outputs.

Challenges & Learnings

One major challenge was maintaining deterministic structure across multiple agent interactions. Without schema validation, outputs became inconsistent and difficult to orchestrate.

To solve this, structured JSON schemas and validation guards were implemented between each agent stage. This significantly improved reliability and reduced cascading errors.

Another challenge was balancing assessment rigor with realistic readiness scoring. Implementing a critic layer helped ensure exam alignment and prevented overly optimistic certification recommendations.

The project reinforced the importance of validation loops, adaptive orchestration, and modular agent design in production AI systems.

Contact Information

https://www.linkedin.com/in/rishisaiteja

Country/Region

India

This is a project submission rather than a scoped change in the agentsleague repository. Review the linked project’s README.md, requirements.txt, .env.example, main.py, and run_evaluation.py first. Verify that the documented setup, main workflow, and evaluation harness run as described; the submission checklist defines the expected completion criteria.