CareerAI - AI Career Guidance System

CareerAI is a full-stack machine learning web application that recommends top career paths from aptitude and performance inputs. It uses a trained Gradient Boosting classifier and serves both API and frontend pages from one Flask service.

Live URL: https://career-ai-wefq.onrender.com/

Repository: https://github.com/suresh4330/Career-ai.git

Table of Contents

• Overview
• Feature Highlights
• Resume Highlights
• Screenshots
• Tech Stack
• Architecture
• Architecture Diagram
• How It Works
• Project Structure
• API Reference
• Input Schema
• Run Locally
• Deploy on Render
• Troubleshooting
• FAQ
• Contributing
• Changelog
• Roadmap
• Author
• License

Overview

CareerAI helps students and early professionals explore likely-fit career paths from structured assessment data. The model returns top-5 ranked recommendations with confidence scores, and the backend exposes clean endpoints for app integration.

Feature Highlights

• End-to-end ML recommendation workflow.
• Top-5 career predictions with confidence percentages.
• 8 aptitude dimensions and 8 performance indicators.
• Feature-engineered input vector aligned with training schema.
• Flask backend serving both API and frontend static pages.
• Render-ready single-service deployment.

Resume Highlights

• Built and deployed a production-style ML recommendation app using Flask and scikit-learn.
• Designed feature engineering pipeline with 25 model-aligned features for robust inference.
• Implemented ranked top-5 prediction endpoint with confidence scoring.
• Delivered one-service deployment architecture where backend serves both API and UI.
• Added operational docs for local run, deployment, and troubleshooting.

Resume-ready one-liner:

Developed and deployed a full-stack AI Career Guidance platform using Flask and Gradient Boosting to generate top-5 career recommendations with confidence scoring from aptitude and performance data.

Screenshots

Add screenshots to:

• docs/screenshots/

Suggested file names:

• docs/screenshots/home.png
• docs/screenshots/features.png
• docs/screenshots/assessment.png
• docs/screenshots/results.png

Markdown snippet:

### Home
![Home](docs/screenshots/home.png)

### Features
![Features](docs/screenshots/features.png)

### Assessment
![Assessment](docs/screenshots/assessment.png)

### Results
![Results](docs/screenshots/results.png)

Tech Stack

Frontend

• HTML5
• Tailwind CSS
• JavaScript

Backend

• Python
• Flask
• Flask-CORS
• Gunicorn

Machine Learning

• scikit-learn
• NumPy
• pandas
• joblib

Architecture

1. User submits aptitude and performance values from UI.
2. Flask API builds engineered feature vector.
3. Gradient Boosting model computes class probabilities.
4. API returns ranked top-5 careers with confidence.
5. Flask serves HTML pages and static assets from same service.

Architecture Diagram

flowchart LR
    UI[HTML Tailwind JS Frontend] --> API[Flask Backend]
    API --> FE[Feature Engineering 25 Features]
    FE --> ML[Gradient Boosting Classifier]
    ML --> OUT[Top 5 Careers + Confidence]
    API --> PAGES[Static Pages + Assets]

Loading

How It Works

• Aptitude input: 8 numeric fields.
• Performance input: 8 categorical fields (POOR, AVG, BEST).
• Engineered features: aggregate score, diversity, intelligence/creativity/social/physical indices, performance score, high-performer flag, optional cluster.
• Inference output: sorted top-5 predictions from model probabilities.

Project Structure

Career-ai/
|- assets/
|  |- css/
|  |- js/
|- backend/
|  |- app.py
|  |- career_prediction_model.joblib
|  |- feature_order.json
|  |- feature_scaler.joblib
|  |- label_encoder.joblib
|  |- model_summary_report.txt
|  |- requirements.txt
|- index.html
|- about.html
|- features.html
|- test.html
|- contact.html
|- Procfile
|- render.yaml
|- requirements.txt
|- README.md

API Reference

GET /api

Returns API metadata and endpoint summary.

GET /health

Returns service status, model-loaded state, and expected feature count.

Example response:

{
  "status": "OK",
  "model_loaded": true,
  "features_expected": 25
}

GET /careers

Returns all available career labels and total count.

POST /predict

Returns top-5 ranked career predictions.

Request example:

{
  "aptitudes": {
    "linguistic": 12,
    "musical": 10,
    "bodily": 11,
    "logical_mathematical": 15,
    "spatial_visualization": 14,
    "interpersonal": 13,
    "intrapersonal": 12,
    "naturalist": 11
  },
  "performance": {
    "project_performance": "AVG",
    "practical_skills": "AVG",
    "research_interest": "AVG",
    "communication_skills": "AVG",
    "leadership_qualities": "AVG",
    "teamwork": "AVG",
    "time_management": "AVG",
    "self_learning": "AVG"
  },
  "cluster": 0
}

Response example:

{
  "status": "success",
  "top_predictions": [
    {
      "rank": 1,
      "career": "Economist",
      "confidence": 31.11
    }
  ]
}

Input Schema

Aptitude keys:

• linguistic
• musical
• bodily
• logical_mathematical
• spatial_visualization
• interpersonal
• intrapersonal
• naturalist

Performance keys:

• project_performance
• practical_skills
• research_interest
• communication_skills
• leadership_qualities
• teamwork
• time_management
• self_learning

Accepted values for performance fields:

• POOR
• AVG
• BEST

Run Locally

git clone https://github.com/ManikantaPerla07/Career-ai
cd Career-ai
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
python backend/app.py

Open:

• http://127.0.0.1:5000

Deploy on Render

This repository is configured for single-service deployment.

Render settings:

• Build command: pip install -r requirements.txt
• Start command: gunicorn backend.app:app
• Environment: Python
• File used: render.yaml

Troubleshooting

If model fails to load:

• Confirm these files exist in backend/:
  • career_prediction_model.joblib
  • label_encoder.joblib
  • feature_order.json

If prediction returns feature mismatch:

• Ensure payload uses all required aptitude and performance keys.
• Confirm no custom code changed feature engineering order.

If static files do not load:

• Verify assets are in assets/ and URL path is /assets/.
• Check that app is started from repository root context.

If Render build fails:

• Confirm requirements.txt includes backend dependencies via -r backend/requirements.txt.

FAQ

Why top predictions instead of one career?

The model outputs class probabilities. Returning top-5 improves usefulness and transparency.

Is this suitable for final career decisions?

No. It is a guidance tool for educational use and should be combined with human mentoring.

Why do confidence values not sum to 100 in response?

Only top-5 predictions are returned; not all classes are shown.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md before opening a pull request.

Changelog

Release updates are tracked in CHANGELOG.md.

Roadmap

• Explainability output for model decisions.
• User login and recommendation history.
• Better analytics dashboard for trend insights.
• Enhanced mobile-first UI refinement.

Author

Suresh Kumar

• Email: careerai.help@gmail.com

License

This project is intended for educational and academic use.