Skip to main content

AWS Serverless YouTube Popularity Predictor

This project is a complete, end-to-end Machine Learning Operations (MLOps) pipeline built on Amazon Web Services (AWS). It trains a model to predict the potential view count of a YouTube video and deploys it as a scalable, cost-effective, serverless API with an interactive web frontend.

The entire architecture is designed to run with zero ongoing monthly costs, leveraging the AWS "Always Free" Tier.

🏆 Key Achievements​

This project showcases a full range of modern MLOps skills, from automated cloud-based builds to a fully serverless, user-facing application.

  • Cost-Effective Architecture: Engineered the entire prediction service and frontend to operate on the AWS Always Free Tier, resulting in $0/month ongoing costs.
  • CI/CD for Machine Learning: Implemented a complete CI/CD pipeline using AWS CodeBuild and GitHub.
  • Serverless First: Deployed the model as a highly-available, auto-scaling serverless function using AWS Lambda and API Gateway.
  • Static Site Hosting: Deployed a responsive frontend as a static website using AWS S3.
  • Containerized Training: Orchestrated model training as a one-off, containerized batch job on AWS ECS Fargate.