Career Compass - AI-powered job matching system

AI-powered job matching system that intelligently connects candidates with opportunities using advanced ML models and workflow automation.

Overview

Career Compass analyzes resumes, extracts skills, and matches candidates with relevant job opportunities using a combination of AI models, workflow orchestration, and intelligent decision-making.

Features

• Resume Analysis - AI-powered skill extraction and profile assessment
• Intelligent Matching - Dynamic job matching based on skills and experience
• Workflow Automation - Automated pipeline orchestration
• Real-time Dashboard - Interactive interface for job discovery

Tech Stack

• Cline - Autonomous coding workflows
• Kestra - Workflow orchestration with AI Agent
• Oumi - ML model training and inference
• CodeRabbit - Automated code quality reviews
• Vercel - Serverless deployment and hosting

Quick Start

Installation

# Install dependencies
pip install -r requirements.txt

Training Model

# Train the Oumi model
python3 train_model.py

Start CLI

# Option 1: Use startup script
./start.sh

# Option 2: Manual start
conda activate hackathon  # if using conda
python3 compass_cli.py

Deploy to Vercel

npx vercel --prod

Project Structure

career-compass/
├── api/                  # Vercel serverless functions
│   ├── analyze.js       # Main integration endpoint
│   ├── kestra.js         # Kestra workflow execution
│   ├── oumi.js           # Model inference
│   ├── coderabbit.js     # Code quality status
│   └── technologies.js   # Technology status
├── compass_cli.py        # CLI interface
├── train_model.py        # Model training script
├── cline_workflow.js     # Cline autonomous workflows
├── index.html            # Web interface
├── kestra_flow.yaml      # Kestra workflow configuration
└── oumi_train.yaml       # Oumi training configuration

API Endpoints

/api/analyze - Combined Analysis (Main Endpoint)

Method: POST
Purpose: Orchestrates all sponsor tools together

Request:

{
  "resume_text": "Resume content here..."
}

Response:

{
  "success": true,
  "integration": {
    "cline": { "skills_extracted": 5, "status": "active" },
    "oumi": { "confidence": 0.89, "profile_quality": "high" },
    "kestra": { "workflow_id": "career_compass_logic", "status": "triggered" },
    "coderabbit": { "active": true, "code_quality": "excellent" },
    "vercel": { "deployed": true, "region": "iad1" }
  },
  "matches": [...]
}

/api/kestra - Workflow Execution

Method: POST
Executes Kestra workflow (fetch data → AI analysis → log results)

/api/oumi - Model Inference

Method: POST
Analyzes job descriptions using trained Oumi model

/api/coderabbit - Code Quality Status

Method: GET
Returns CodeRabbit integration status and metrics

/api/technologies - Technology Status

Method: GET
Returns status of all integrated technologies

Award Criteria Fulfillment

✅ Cline CLI Award

Criterion: Best project using Cline CLI for autonomous coding workflows

Evidence:

• cline_workflow.js with 4 autonomous workflows:
  • Code generation (generate)
  • Automated testing (test)
  • Code review (review)
  • Deployment (deploy)

✅ Kestra AI Agent Award

Criterion: Using Kestra's AI Agent to summarize data and make decisions

Evidence:

• kestra_flow.yaml implements:
  • Data summarization from multiple sources
  • Decision-making based on summarized data
  • Structured JSON output with recommendations

✅ Oumi Award

Criterion: Creative use of Oumi to train/evaluate LLMs and OSS contributions

Evidence:

• Trained custom model (SmolLM2-135M-Instruct) for job quality assessment
• Training config: oumi_train.yaml
• OSS contribution: oumi_contribution.md with reusable templates
• Model saved: output/salary_model/

✅ CodeRabbit Award

Criterion: Best open-source engineering with clean PRs and OSS workflows

Evidence:

• .coderabbit.yaml configuration
• .github/PULL_REQUEST_TEMPLATE.md for standardized PRs
• .github/workflows/coderabbit.yml for automated reviews
• CONTRIBUTING.md with OSS guidelines

Security

Environment Variables

Never commit API keys. Use .env file (excluded from git):

# Copy example file
cp .env.example .env

# Add your keys
GROQ_API_KEY=your_actual_key_here

Protected Files

• .env, .env.local, *.env - Never commit
• output/ - Large model files excluded
• .venv/ - Virtual environment excluded

Best Practices

• ✅ Use environment variables for all secrets
• ✅ Never hardcode API keys
• ✅ Rotate keys if accidentally exposed
• ✅ Use different keys for dev/prod

Contributing

Pull Request Process

1. Fork the repository
2. Create feature branch: git checkout -b feature/amazing-feature
3. Make changes and ensure CodeRabbit passes
4. Commit: git commit -m 'Add amazing feature'
5. Push: git push origin feature/amazing-feature
6. Open Pull Request

CodeRabbit Integration

All PRs are automatically reviewed for:

• Code quality and best practices
• Security vulnerabilities
• Documentation completeness
• Performance optimizations

Code Style

• Follow existing code style
• Add comments for complex logic
• Include docstrings for functions
• Update tests when fixing bugs

Cline Workflows

Code Generation

node cline_workflow.js generate "API endpoint, database model, tests"

Automated Testing

node cline_workflow.js test .

Code Review

node cline_workflow.js review src/api.js

Deployment

node cline_workflow.js deploy . vercel

Kestra Workflow

The kestra_flow.yaml defines a 3-task workflow:

1. Fetch Data - Retrieves job data from external sources
2. AI Analysis - Uses Groq AI to analyze and summarize data
3. Log Result - Outputs final results

Oumi Model Training

Configuration

• Model: SmolLM2-135M-Instruct
• Task: Job quality assessment and salary evaluation
• Config: oumi_train.yaml
• Output: output/salary_model/

Training

python3 train_model.py

OSS Contribution

See oumi_contribution.md for reusable configuration templates and best practices shared with the Oumi community.

Deployment

Vercel

npx vercel --prod

The api/ directory contains Vercel Serverless Functions that automatically deploy as endpoints.

Environment Variables (Vercel)

Set in Vercel dashboard:

• GROQ_API_KEY - For Kestra AI Agent workflows

License

MIT

Acknowledgments

• WeMakeDevs - Special thanks for offering this opportunity to work on the AI Agents Assemble 2025 hackathon
• Cline for autonomous code generation
• Kestra for workflow orchestration
• Oumi for custom model training
• Vercel for seamless deployment
• CodeRabbit for code quality assurance