MLOps-Project

Repository contains end to end mlops project with detailed tutorial and code walkthroughs

Project components:

1. Data Loading

• Uses Yellow Taxi Trip data from NYC in parquet format
• Data loading implemented in load_data function
• Loads trip data including pickup/dropoff locations, distances, and timestamps

2. Model Training Pipeline

• Feature engineering and preprocessing in preprocess_data
• Data standardization using scikit-learn's StandardScaler
• Linear Regression model to predict trip duration
• Model evaluation using MSE metric
• Model and scaler artifacts saved using joblib

3. Model Inference Pipeline

• FastAPI web service for model serving
• REST API endpoint for trip duration predictions
• Takes pickup location, dropoff location, and trip distance as inputs
• Returns predicted trip duration

4. MLflow Tracking

• Experiment tracking and model versioning
• Logs model parameters, metrics, and artifacts
• Model registry for managing different versions
• UI available at http://127.0.0.1:5000

5. Workflow Orchestration

• Uses Prefect for pipeline orchestration
• Task dependencies and retries configured
• Handles data loading, preprocessing, training and evaluation
• Automated model artifact management

Prefect Runs Image: