The recruiter was very nice. The technical interview consisted of a 1-hour live coding session focused on training a model for their data in PyTorch.

There were 10 questions that required you to write or complete code and explain your answers in writing; most of them were quite easy.

The challenge lies in completing all of them within an hour.

While 1 hour for all the tasks is unrealistic, it is possible to complete many if you code very quickly.

No tools are available to you during the exercise other than library references; Stack Overflow, autocomplete, or coding assistants are not allowed.

The knowledge tested felt too basic, more like data science rather than machine learning, and the techniques are old, unlike any foundational models we might work with these days.

My only feedback is that it would be better to go through a harder exercise that actually tests modern deep learning, but with more time.

My general experience was good, and the interviewer did a great job moderating the exercise.

Load CSV into pandas and describe sizes, nature of variables, etc.