So, let’s suppose we’re building an education app for students.

Our app needs to do three things:

  1. test the user with past exam questions, which then must be marked
  2. answer any question the user asks about their subject
  3. ask the user new questions, which must also be marked

We will be synthesizing the data for demonstration purposes, such that you can run everything locally.

You can jump straight to any of the following sections, but we would suggest going in the following order: Logging, Evaluations, then Debugging.

If you’re interested how the various synthetic data for the case study was generated, you can check out our data generation notebook, but this is not necessary for completing the case study. We’ll simply be loading this pre-generated data from a data store in each of the examples.