Course Wiki

Serverless Deployment

Serving models from AWS Lambda functions instead of a persistent server, exposed through API Gateway.

Module 9 deploys models with AWS Lambda: the model and its dependencies are packaged into a Lambda-compatible container or layer, the function loads the model on invocation, and API Gateway exposes it over HTTP. The pay-per-request model fits spiky or low-traffic prediction loads without a server to maintain.

The engineering work is packaging: TensorFlow and PyTorch dependencies exceed Lambda size limits, so the course builds Lambda containers with the model baked in, and shows TensorFlow Serving as the alternative path. Learners leave with a prediction endpoint that scales to zero.

Taught in

• Machine Learning Zoomcamp

  • Module 9: Serverless Deep Learning
    • Introduction to Serverless
    • AWS Lambda
    • Creating the lambda function
    • API Gateway: exposing the lambda function
    • Explore more
    • Python 3.12 vs TF Lite 2.17 ## Related concepts

• Kubernetes

• FastAPI

• Docker

• Neural Networks

• Transfer Learning