README.md

metadata

title: AutoML MLOps Pipeline
emoji: 🤖
colorFrom: blue
colorTo: green
sdk: docker
app_port: 8000
pinned: false
license: mit

🤖 AutoML MLOps Pipeline

Production-ready end-to-end AutoML pipeline with MLflow tracking, comprehensive monitoring, and automated orchestration.

🚀 Features

• 🤖 AutoML: AutoGluon, FLAML, PyCaret integration
• 📊 MLflow Tracking: DagsHub integration with comprehensive metrics
• 🔍 Monitoring: Drift detection, prediction logging, performance tracking
• 📈 Observability: Prometheus metrics & Grafana dashboards
• 🔄 Orchestration: Airflow DAGs for automated scheduling
• 🐳 Docker: Complete containerization with docker-compose
• ⚡ FastAPI: RESTful API with 11+ endpoints
• 🎯 CI/CD: GitHub Actions for automated testing and deployment

📋 Pipeline Stages

1. Data Ingestion - Load and validate dataset
2. Data Validation - Schema validation and quality checks
3. Data Transformation - Feature engineering and preprocessing
4. AutoML Training - Multi-framework model training
5. Model Evaluation - Comprehensive metrics and validation
6. Model Comparison - Best model selection
7. Model Pusher - Production model deployment

🛠️ Tech Stack

• ML Frameworks: AutoGluon, FLAML, PyCaret
• API: FastAPI, Uvicorn
• Tracking: MLflow, DagsHub
• Monitoring: Prometheus, Grafana, Evidently AI
• Orchestration: Apache Airflow
• Containerization: Docker, Docker Compose
• CI/CD: GitHub Actions

📦 Quick Start

Local Development

# Clone repository
git clone https://github.com/Abeshith/AutoML-MLOps-PipeLine.git
cd AutoML-MLOps-PipeLine

# Create virtual environment
python -m venv automlenv
source automlenv/bin/activate  # On Windows: automlenv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Set environment variables
cp .env.example .env
# Edit .env with your credentials

# Run training pipeline
python scripts/train.py

# Start API server
python scripts/serve.py --reload

Docker Deployment

# Start all services
docker-compose up -d

# Access services
# API: http://localhost:8000/docs
# Prometheus: http://localhost:9090
# Grafana: http://localhost:3000 (admin/admin)

🌐 API Endpoints

Prediction

POST /predict
{
  "age": 45,
  "sex": 1,
  "cp": 2,
  "trestbps": 130,
  "chol": 250,
  "fbs": 0,
  "restecg": 1,
  "thalach": 150,
  "exang": 0,
  "oldpeak": 2.5,
  "slope": 2,
  "ca": 0,
  "thal": 2
}

Training

POST /train
GET /train/status

Monitoring

GET /monitoring/metrics          # Prometheus metrics
GET /monitoring/health/drift     # Drift detection status
GET /monitoring/performance/summary
GET /monitoring/reports/daily

📊 Model Performance

• Validation Accuracy: 88.84%
• Test Accuracy: 88.68%
• ROC-AUC: 95.48%
• Best Model: WeightedEnsemble_L3

🔧 Utility Scripts

# Train model
python scripts/train.py

# Evaluate model
python scripts/evaluate.py --model-path <path>

# Start API server
python scripts/serve.py --host 0.0.0.0 --port 8000 --reload

# Initialize Airflow
python scripts/init_db.py

🔄 Airflow Orchestration

# Set AIRFLOW_HOME
export AIRFLOW_HOME=$(pwd)/airflow

# Initialize database
python scripts/init_db.py

# Start services
airflow scheduler  # Terminal 1
airflow webserver  # Terminal 2

# Access UI: http://localhost:8080

📈 Monitoring Stack

• Drift Detection: KS test for numerical features
• Prediction Logging: JSONL format with threading
• Performance Tracking: Batch-level metrics
• Report Generation: Daily/weekly JSON reports
• Prometheus Metrics: Request count, latency, accuracy, drift status
• Grafana Dashboards: 5-panel visualization

🐳 Docker Services

• FastAPI App (8000): Main ML API
• Prometheus (9090): Metrics collection
• Grafana (3000): Visualization dashboards

🔐 Environment Variables

MLFLOW_TRACKING_URI=your_dagshub_uri
DAGSHUB_TOKEN=your_token

📚 Documentation

• Docker Setup
• Scripts Usage
• CI/CD Workflows
• Airflow Guide

🧪 CI/CD Pipeline

Automated Workflows

• CI: Lint with flake8, format check with black
• Docker Build: Build and push to GitHub Container Registry
• HuggingFace Deploy: Auto-deploy to Spaces on push

Container Images

docker pull ghcr.io/abeshith/automl-mlops-pipeline:latest

📊 Project Structure

AutoML-MLOps-PipeLine/
├── src/mlpipeline/          # Core pipeline components
├── app/                      # FastAPI application
├── config/                   # Configuration files
├── scripts/                  # Utility scripts
├── airflow/                  # Airflow DAGs
├── monitoring/               # Monitoring components
├── observability/            # Prometheus/Grafana configs
├── notebooks/                # Jupyter notebooks
├── Dockerfile                # Container definition
├── docker-compose.yaml       # Multi-service orchestration
└── requirements.txt          # Python dependencies

⭐ Star this repo if you find it helpful!