Model Deployment for AI Engineers
For engineers who have a trained model and need to turn it into a reliable, secure service. You will export it to a portable format, serve it with FastAPI in a lean container, make it smaller and faster, release it safely with canaries and A/B tests, and harden, monitor and automate it through CI/CD.
Course Content
Turn a trained model into a portable artefact, serve it correctly with FastAPI or Flask, and package it in a small, secure, fast-building Docker image. 3 lessons, about 45 minutes.
Shrink and speed up a model with quantization, pruning and distillation, release new versions safely with canaries and A/B tests, and instrument the service so you can see what it is doing. 3 lessons, about 45 minutes.
Choose the right cloud target by traffic shape and cost, build a CI/CD pipeline that tests the model as well as the code, and harden the API against abuse and model theft. 3 lessons, about 45 minutes.
Take a CIFAR-10 classifier from training to a tested, containerised REST API with CI and a deployment that someone else can run. 1 lesson, about 15 minutes.