Work

End-to-End ML Deployment Pipeline

MLOps
Flask
Docker
CI/CD
Python

A fully automated ML deployment pipeline predicting California housing prices — Flask web app, Docker containerization, and GitHub Actions CI/CD pushing to Heroku on every commit with zero manual intervention.

Overview

Engineered a production-grade machine learning deployment pipeline predicting California housing prices using linear regression, complete with automated CI/CD infrastructure. Built a professional Flask web interface with a custom dark-themed design, containerized the application with Docker for environment portability, and wired up GitHub Actions to automatically build, containerize, and deploy to Heroku on every push to main — zero manual steps required.

GitHub: saideepa05/housing_deployment

What I Built

  • Flask web application — responsive dark-themed UI with a /predict REST endpoint, strict numeric input validation, and explicit feature mapping to the trained model
  • Linear regression model — retrained with scikit-learn to resolve version compatibility issues that would break inference in production
  • Docker containerization — full Dockerfile packaging the app, model, and dependencies for consistent behaviour across development and production environments
  • GitHub Actions CI/CD pipeline — automated workflow triggered on push to main: builds the container, authenticates with Heroku CLI, and deploys with zero-downtime releases
  • Infrastructure debugging — resolved scikit-learn version mismatches and Heroku CLI ENOENT spawn errors that blocked the initial pipeline, resulting in a fully reliable deploy chain

Tech Stack

Python · Flask · scikit-learn · Pandas · NumPy · Docker · GitHub Actions · Heroku · HTML/CSS

Results

Delivered a fully automated deployment pipeline with zero manual intervention — a push to main triggers containerization and production release end-to-end. Resolved critical infrastructure blockers at every layer (model versioning, container builds, Heroku CLI integration) and implemented robust input validation, resulting in a production-ready application serving real-time housing price predictions.