There are three stages for sure, and potentially a fourth:

• A half-hour HR screen, which went smoothly.

• A test case where you must build a "production grade" API with a machine learning model behind it. This stage is why I'm rating my overall experience as negative. They claim the allotted time for solving is four days, but unfortunately, to be exhaustive with the details and make it "production grade" takes much more than that. Some companies are now assigning very time-consuming projects which candidates are expected to complete as if they had nothing else going on in their lives. The worst part? I received zero feedback and zero questions about the API. Quite annoying, right?

• A two-hour technical interview with the Machine Learning team, which was quite fair. They asked general questions around Machine Learning and had me suggest improvements to a Python module.

There may be a fourth stage, which I did not reach.

1. Build a production-ready API with an ML model behind it.

2. Explain the different optimization algorithms in Deep Neural Networks (DNNs).

3. What are kernels, and what is their use?

4. Why are Convolutional Neural Networks (CNNs) translation invariant?

5. Review a Python module and suggest changes.

6. How would you regularize Random Forests?

7. How to make linear regression and Random Forests robust to outliers?