Certway AI

Certway AI is a local Streamlit prototype for a Microsoft Foundry-ready multi-agent certification coach. It demonstrates grounded, synthetic-data reasoning for enterprise learning and workforce readiness.

Azure AI Foundry status: this project is currently a local prototype / Foundry-ready placeholder. It does not make live Azure model calls yet. No API Keys, Credentials, Paid Services, Confidential Data, or Real Employee Data are required for the demo.

The project helps employees prepare for role-based certifications by recommending learning paths, generating workload-aware study plans, creating practice assessments, estimating readiness risk, and giving managers team-level insights.

Demo

Demo video : https://drive.google.com/file/d/12N4C69H54f1Duj0LvJgq0W4t-r07iPx6/view?usp=sharing

Repository: https://github.com/Hasnainali19/certway-ai

Hackathon Track

Agents League Hackathon — Reasoning Agents Track

This project is designed for the Reasoning Agents challenge and demonstrates:

• Multi-agent system design
• Role-based certification planning
• Workload-aware study scheduling
• Grounded practice assessment using synthetic knowledge documents
• Manager-level readiness insights
• Safe synthetic data usage
• Microsoft Foundry / Azure AI Foundry-ready placeholder architecture

Problem

Organizations often struggle to manage internal certification programs because:

• Employees do not know what to study first.
• Study plans are not personalized to role or workload.
• Managers lack visibility into certification readiness.
• Practice questions may be generic or ungrounded.
• Learning progress is difficult to track across teams.

Solution

Certway AI uses multiple specialized agents to support employees and managers throughout the certification preparation process.

Project Architecture

Layer	Files	Purpose
Streamlit app and UI pages	app.py	Provides Overview, Learner Coach, Practice Assessment, and Manager Dashboard pages.
Synthetic learner data	data/learners.csv	Provides synthetic learner profiles, certification goals, practice scores, and study progress.
Synthetic workload data	data/workload_signals.csv	Adds meeting load, focus capacity, and preferred learning slot signals.
Certification metadata	data/certifications.csv	Stores role-to-certification mappings, recommended study hours, and readiness thresholds.
Foundry IQ-style grounding source	docs/engineering_certification_guide.md	Acts as the approved synthetic guide used to extract grounded topics and source sections.
Agent logic	Agent helper functions in app.py	Implements local Orchestrator, Learning Path, Readiness, Assessment, Study Plan, Engagement, and Manager Insights behavior.
Agent prompt documentation	prompts/agent_prompts.md	Documents intended agent roles, responsibilities, and safety rules.
Azure placeholder client	utils/azure_client.py	Provides local recommendation fallback and honest Foundry placeholder/config status. It does not call Azure models.

Multi-Agent Architecture

1. Orchestrator Agent

Receives learner or manager requests and coordinates the workflow across specialized agents.

2. Learning Path Curator Agent

Maps a learner's role and certification goal to relevant learning topics.

3. Study Plan Generator Agent

Creates a realistic study plan based on practice score, hours studied, focus time, and meeting load.

4. Assessment Agent

Generates practice questions and readiness recommendations based on approved synthetic certification content.

5. Engagement Agent

Suggests study timing and support strategies based on workload signals.

6. Manager Insights Agent

Summarizes team-level readiness, risk areas, and recommended manager actions.

Microsoft IQ / Foundry Alignment

This prototype uses a synthetic certification guide as an approved knowledge source to demonstrate a Foundry IQ-style grounding pattern.

docs/engineering_certification_guide.md

The Streamlit app parses this guide by role and certification, then shows grounded topics and source sections in the learning path and practice assessment.

Azure AI Foundry Integration

The project includes an Azure AI Foundry-ready placeholder client:

utils/azure_client.py

Current mode:

• Local rule-based multi-agent prototype
• Azure AI Foundry / Microsoft Foundry placeholder only
• Safe fallback if Azure is not configured
• Environment-based setup using .env.example
• No live Azure model calls are made

The environment template is provided in:

.env.example

Real Azure values should be stored in a local .env file and should never be committed to GitHub.

Reasoning Flow

1. Load synthetic learner, workload, and certification metadata from CSV files.
2. Parse the approved synthetic certification guide by role and certification.
3. Select grounded learning topics and source sections for the learner's target certification.
4. Compute readiness signals from practice score, study progress, meeting load, and focus capacity.
5. Route the profile through the local multi-agent workflow:
   • Orchestrator Agent
   • Learning Path Curator Agent
   • Readiness Reasoning Agent
   • Assessment Agent
   • Study Plan Generator Agent
   • Engagement Agent
   • Manager Insights Agent
6. Present a judge-facing final recommendation that combines readiness stance, key evidence, grounded topics, learner action, and manager handoff.
7. Keep all recommendations framed as coaching support, not a guarantee of certification success.

Judging Criteria Mapping

Hackathon criterion	How Certway AI addresses it
Accuracy & Relevance	Uses role, certification, practice score, study progress, workload signals, and certification-specific readiness thresholds.
Reasoning & Multi-step Thinking	Shows readiness signals, agent responsibilities, evidence, decisions, and a final recommendation summary.
Creativity & Originality	Combines certification coaching with workload-aware study planning and manager-ready support guidance.
User Experience & Presentation	Provides a polished Streamlit UI with Overview, Learner Coach, Practice Assessment, and Manager Dashboard pages.
Reliability & Safety	Uses synthetic data only, avoids secrets, includes human review language, and keeps Azure status honest.
Foundry IQ-style grounding	Parses an approved synthetic guide and displays grounded topics and source sections for practice questions.
Clear demo readiness	Includes a simple local setup path and a demo script for judges.

Synthetic Data

This project uses synthetic demo data only.

Datasets:

data/learners.csv
data/workload_signals.csv
data/certifications.csv

No real employee data, customer data, credentials, confidential data, or personally identifiable information is used.

Tech Stack

• Python
• Streamlit
• Pandas
• Plotly
• Azure AI Foundry-ready placeholder architecture
• GitHub
• GitHub Copilot
• VS Code

How to Run Locally

1. Clone the repository

git clone https://github.com/Hasnainali19/certway-ai.git
cd certway-ai

2. Create a virtual environment

python -m venv .venv

3. Activate the virtual environment

On Windows:

.venv\Scripts\activate

On macOS/Linux:

source .venv/bin/activate

4. Install dependencies

pip install -r requirements.txt

5. Run the Streamlit app

streamlit run app.py

Environment Variables

Copy .env.example to .env if using Azure AI Foundry.

On Windows:

copy .env.example .env

On macOS/Linux:

cp .env.example .env

Then update the .env file with your real Azure values.

Do not commit .env.

Responsible AI and Safety

Certway AI follows these safety practices:

• Uses synthetic data only
• Avoids real employee or customer data
• Does not expose credentials or secrets
• Provides readiness recommendations, not guaranteed outcomes
• Encourages human review for important workforce decisions
• Clearly separates local prototype mode from Azure AI Foundry mode

GitHub Copilot Usage

GitHub Copilot was used during development to support:

• Streamlit UI generation
• Python helper function creation
• Multi-agent workflow documentation
• Debugging and code explanation
• README drafting and refinement

Demo Script

A sample judge demo flow:

1. Open the Streamlit app.
2. View the overview dashboard.
3. Open Learner Coach and select a learner such as Taylor Kim or Alex Morgan.
4. Review the readiness stance and the judge-facing final recommendation card.
5. Open the Reasoning Trace tab and review each agent's responsibility, evidence, and decision.
6. Open the Study & Engagement Plan tab to see workload-aware study guidance.
7. Open the Foundry Recommendation tab and note that it is a local / Foundry-ready placeholder, not a live Azure call.
8. Open Practice Assessment and review grounded questions with source sections from the synthetic guide.
9. Open Manager Dashboard and review team-level readiness and manager action guidance.

Project Status

Current version:

• Working Streamlit MVP
• Synthetic datasets added
• Multi-agent reasoning trace with responsibilities, evidence, and decisions
• Grounded knowledge source parsing and visible source sections
• Azure AI Foundry placeholder status with no live Azure model calls
• Ready for further Microsoft Foundry integration

License

MIT License