Build a Smart House Price Prediction system using ML, deployed via FastAPI, containerized with Docker, and hosted on AWS App Runner with CI/CD automation.

Smart House Price Prediction (MLOps Project)

A complete end-to-end MLOps project demonstrating how to build, containerize, and deploy a machine learning model using FastAPI, Docker, and AWS App Runner. This project predicts house prices in different locations of bangalore based on features like BHK, square footage, location, and amenities.

Project Overview

This project showcases:

• Machine Learning model (RandomForestRegressor)
• FastAPI backend with REST API + UI
• Docker containerization
• Deployment on AWS (ECR + App Runner)
• CI/CD automation with GitHub Actions
• Monitoring via CloudWatch

Architecture Flow

User → FastAPI → ML Model → Docker → ECR → App Runner → CI/CD → CloudWatch

Getting Started

1. Clone Repository

git clone https://github.com/disisdsid/mlops-project-1 cd mlops-project-1

2. Train Model

python train_house.py This generates house_model_v1.pkl used by the FastAPI app.

3. Run Locally

uvicorn main_house:app --reload Visit:

http://127.0.0.1:8000/ → Health check

http://127.0.0.1:8000/ui → Web UI

http://127.0.0.1:8000/predict → API endpoint

http://127.0.0.1:8000/docs → Swagger UI

4. Docker Build

docker build -t house-price-api .

5. Push to AWS ECR

aws ecr create-repository --repository-name house-price-api docker tag house-price-api:latest <ECR_URI> docker push <ECR_URI>

6. Deploy on AWS App Runner Select ECR image

Set port to 8000

Configure health check (recommended: HTTP /health)

Key API Endpoints / → Health check

/ui → User Interface form

/predict → JSON API prediction

/docs → Swagger UI

CI/CD Pipeline GitHub Actions automates:

Code push

Docker build

Push to ECR

Deploy to App Runner

Challenges Faced Docker daemon issues

Setting up UI with python code

UI route confusion (/ui)

Improvements & Future Work Add database integration (DynamoDB)

Use larger real-world dataset

Model versioning with S3

Setup custom domain

Conclusion This project demonstrates a complete MLOps lifecycle — from model training to cloud deployment — and is suitable for beginners to understand real-world ML deployment pipelines.

License This project is licensed under the MIT License.

Created by Siddhanta Dash

LinkedIn - https://www.linkedin.com/in/siddhanta-dash

Note: I have removed the security credentials after testing so don't try to use them as they won't work.. LOL...