A Real Estate company have data on some properties i.e house. Data like the number of bedrooms, size of the front porch etc. They want a model that predicts the price of the house given the house features. This is a regression task. The machine learning engineer (me) sourced data from Kaggle. The dataset contains over 100 columns of different features of houses. He cleaned the data and trained an xgboost model on it.

Cloud Infrastructure:

Run this command to setup the python environment and pre-commit

make setup

MLflow was used for experiment traccking and model registering.

To start up the mlflow server, run

bash orchestration/server.sh

keep this terminal open for the training pipeline

Prefect was used for workflow orchestration to get the data, preprocess it, run hyperparameter tuning with hyperopt and train the xbgoost model.

To start the training pipeline:

bash orchestration/train.sh

Checkout from develop branch to ci-cd branch

git checkout -b ci-cd develop

Change the infrastructure production variables to your choice.

Commit and push the changes to the ci-cd remote branch Make a pull request and merge with develop branch

terraform init -backend-config="key=mlops-zoomcamp-prod.tfstate" -reconfigure
terraform output rest_api_url

Change the url in the deployment test file to the rest api url