The first stage was a discussion with the recruiter followed by a pair programming session with 2 engineers. We then ended the session discussing how we would deploy our code into a production environment - this aspect assesses MLOps concepts and software engineering practices.

The process began with a 30-minute screening call with a recruiter covering basic software engineering and machine learning questions. The second stage was a 90-minute live coding exercise with two engineers, focused on parsing and validating a spreadsheet using Python and Pandas. While they explicitly stated it was an open-book exercise where I could use any tools or documentation, I was later told that referring to the documentation (which I did once) was one of the main reasons for rejection—a contradictory and frustrating piece of feedback. The interviewers seemed to have a single "correct" solution in mind rather than being open to alternative approaches, despite there being several valid ways to solve the problem