BDA-602 - Machine Learning Engineering

Dr. Julien Pierret

Lecture 13

In Summary

  • HTTP is what we want to use to interface with our model
    • GET requests
  • We want our service to be RESTful
  • FastAPI is what we'll use to wrap our models into a service
  • pickle is what you'll use to serialize your models
    • Could also use dill
  • CURL used to test service via the command line
  • Postman good if you want a GUI application
  • ngrok good for testing the service

No "official" Homework

  • Start on your final project
    • Start building features (predictors) on your dataset
      • This is what I really want you to focus on
      • Build a ton of features based off your data
      • Imagination!!
    • Start analyzing those features with the tools with built in class
      • How do they perform?
      • Can you combine them with other features?
      • Are they correlated with other ones?
        • Is one better than the other? Why?