MLops House Prediction Price

? Project Description

This project demonstrates a production-ready MLOps pipeline that integrates data processing, model training, versioning, deployment, and monitoring using modern cloud-native tools.

? 1. Data Layer (Storage & Exploration)

Data is stored in Amazon S3.

A data scientist performs EDA (Exploratory Data Analysis) using Jupyter notebooks (explore.ipynb).

This stage ensures data understanding and preprocessing before training.

? 2. Machine Learning Pipeline

ML scripts handle:

Feature Engineering (Feature Store)

Model Training Pipeline

Model Versioning (Model Registry)

The trained model is saved as an artifact (e.g., model.pkl).

? 3. Experiment Tracking & Serving

MLflow is used for:

Experiment tracking

Model version control

FastAPI is used to expose the model as an API for inference.

? 4. Application Layer

Streamlit provides a simple UI for interacting with the model.

Users can input data and get predictions through a web interface.

? 5. Containerization & Registry

Applications are containerized using Docker.

Images are stored in Amazon ECR.

? 6. CI/CD & GitOps

Source code is managed in GitHub.

CI/CD pipelines automate:

Build

Test

Deployment

Argo CD is used for GitOps-based deployment.

? 7. Deployment (Cloud Infrastructure)

The application is deployed on Amazon EKS.

Kubernetes ensures:

Scalability

High availability

Container orchestration

? 8. Monitoring & Observability

Prometheus collects metrics.

Grafana visualizes performance dashboards.

This ensures system health and model performance tracking.

? Overall Flow

Data → stored in S3 → analyzed in notebooks

Training pipeline → model saved & tracked in MLflow

Model → served via FastAPI → UI via Streamlit

Docker images → pushed to ECR

CI/CD → deployed via ArgoCD to EKS

Monitoring → Prometheus & Grafana

? Key Value

This project showcases:

End-to-end MLOps lifecycle

Cloud-native deployment on AWS

Automation using CI/CD + GitOps

Scalable and production-grade architecture