They use TestGorilla as an assessment platform and ask everything about tensor slicing and think they can derive someone´s Pytorch skills only based on that. Besides the specific Pytorch part there were some general "intelligence test" - like questions you can practice when having a look at TestGorilla before.

After applying on their website, they contacted me in a week or so. They sent an "assessment test" from TestGorilla. The test consisted of "critical thinking", "problem solving", "pytorch" and "data science". Example questions can be found on Testgorilla's website. Example question for "critical thinking" is that we have premise 1 and premise 2, and which of the options contradicts them. An example of problem solving is to give you a table of weekdays that some people work in the company, and say that Alex cannot work with Joe and Mary on the same day, which days can Alex work? The Pytorch questions were about PyTorch classes and commands that could be used for specific tasks like padding -- 10min for 16 questions. The data science part was about general intuitions on data science topics (20 questions, 10 min). In general, the test was very fast-paced -- each question had on average 30-40 seconds. I invite people to check Testgorilla for the example questions and assess whether this type of assessment is appropriate for hiring senior machine learning scientists. There was also a 1-minute video recording part where one was asked to introduce oneself and say why one is appropriate for this role. Also, a free text (with limited time to fill in, 5min or so), to write when was the last time you used data to draw some conclusions.