OpenAI Interview Loops
OpenAI's interviews differ by role more than most companies' do. Data science runs through a 48-hour take-home, where AI tools are encouraged, and a review that tests whether you can catch errors in AI-written code. Research engineering is the frontier lab loop with the most coding. It has ML and general coding graded on tests, an ML and statistics round that reaches information theory, transformer debugging and a project presentation. What they share is that they're practical. The work you're asked to do looks like the job, and teams shape the content. Data science alone is hired separately for Product, Integrity, Forecasting, Support and User Operations.