Skip to main content

🏛️ Project Architecture

This project is composed of several key AWS services working together in two main phases: a CI/CD and Model Training pipeline, and a Serverless Inference pipeline. The diagram below illustrates the complete end-to-end workflow.

Component Breakdown​

  • Code: Source code (lambda_function.py, train.py) and configuration (Dockerfile, buildspec.yml) is stored in GitHub.
  • Storage: The dataset and trained model artifact (.joblib file) are stored in AWS S3.
  • Model Training: A containerized script is run as a one-off task on AWS ECS Fargate to train the model and save the artifact to S3.
  • CI/CD Pipeline: A push to GitHub triggers AWS CodeBuild, which builds a container image for the Lambda function and pushes it to Amazon ECR.
  • Serverless Inference: AWS Lambda runs the container from ECR, serving predictions.
  • API Layer: Amazon API Gateway provides a public HTTP endpoint that triggers the Lambda function.
  • Frontend: A static website (HTML/CSS/JS) is hosted in an AWS S3 bucket.