Project: Reasoning Agents - CareerCompass AI

[AryaBadugu]

Track

Reasoning Agents (Azure AI Foundry)

Project Name

CareerCompass AI

GitHub Username

@AryaBadugu

Repository URL

https://github.com/AryaBadugu/careercompass-ai

Project Description

CareerCompass AI is a multi-step reasoning career guidance agent that transforms a simple text profile into a comprehensive, data-backed career strategy. Powered by Azure AI Foundry, Azure OpenAI (GPT-4.1-mini), and Azure AI Search, it executes a transparent 6-step reasoning pipeline:

1. Profile Analysis - Extracts skills, education, goals, and experience from free-text using structured prompt engineering.
2. Career Path Search - Queries an Azure AI Search index containing O*NET career data to find verified alignments, preventing hallucinations.
3. Skills Gap Analysis - Compares user skills against career requirements to identify gaps.
4. Opportunity Ranking - Scores career paths (1-10) based on fit, growth, and entry effort.
5. Roadmap Creation - Generates a structured 30/60/90-day learning plan.
6. Action Plan Synthesis - Compiles a polished, markdown career strategy.

Key Features:

• Live Market Demand Intelligence: Queries Himalayas, RemoteOK, Remotive, and GitHub APIs in real-time to calculate demand scores and show active job postings.
• Side-by-Side Career Comparison: Generates a JSON comparison matrix between two career paths.
• Adaptive Feedback Loop: Re-ranks career recommendations based on user feedback.
• Export Deliverables: Generates ReportLab PDF reports and CSV skills checklists.
• Premium UI: Smooth micro-animations and custom glassmorphic layout.

Demo Video or Screenshots

Demo Video: https://youtu.be/hW6J1EBaKXs
Live Demo: https://careercompass-agent.vercel.app/

Architecture Diagram:

Primary Programming Language

Python

Key Technologies Used

• Azure AI Foundry (Project Orchestration)
• Azure OpenAI Service (GPT-4.1-mini)
• Azure AI Search (Knowledge Base RAG)
• Python 3.11 & FastAPI
• React 18 (Vanilla CSS Glassmorphic UI)
• ReportLab (Dynamic PDF Engine)
• Pandas & OpenPyXL

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 Repo: git clone https://github.com/AryaBadugu/careercompass-ai.git
2. Backend Setup: Create virtual env, install dependencies (pip install -r requirements.txt), and configure the .env file with Azure credentials.
3. Ingest Database: Run python setup_knowledge.py to index O*NET careers database into Azure AI Search.
4. Launch Backend: Run uvicorn main:app --reload (FastAPI starts on port 8000).
5. Launch Frontend: Run cd frontend && npm install && npm start (React starts on port 3000).

Technical Highlights

• O*NET RAG Grounding: Integrated Azure AI Search to dynamically ground the reasoning model in a comprehensive O*NET dataset, completely eliminating career-path hallucinations.
• Concurrent API Aggregation: Implemented Python asyncio to query 5 different remote job APIs in parallel, keeping response times under 2 seconds while fetching real-time market metrics.
• Transparent Reasoning UI: Built a custom React state machine to track and display each step of the agent's internal analysis process dynamically as it happens.
• ReportLab PDF Generation: Designed a custom server-side layout engine that compiles the complex JSON output of the 6-step reasoning analysis into a formatted PDF career report.

Challenges & Learnings

• Challenge: Managing response latency while fetching live job data from multiple public APIs simultaneously.
• Solution: Utilized asynchronous HTTP requests in Python to parallelize fetches, parsing, and formatting, reducing execution times from over 12 seconds down to less than 2 seconds.
• Learning: Learned how to design prompts that yield highly structured, reliable JSON schemas from OpenAI, which allowed us to build robust comparison grids and dynamic timelines on the React frontend.

Contact Information

aryabadugu@gmail.com or linkedin.com/in/arya-badugu

Country/Region

India