Neuralink Interview Questions
& Process

The interview process was a bit strange. It started with a 30-minute call with the hiring manager. Then, there was a technical challenge with 3 questions that I had to finish within a week. After they initially rejected me, they surprisingly called me back for a third stage, which was a technical presentation. During the presentation and Q&A, the team of 5 materials engineers seemed disengaged, busy with their own work, and didn't show much reaction. It felt like they were unsure about what they were looking for in a candidate, especially since they are physics-focused but hiring someone for computational simulations using commercial software. I'd think twice about this team.

The interview process involved a single round that I didn't pass. The questions were surprisingly easy, almost trivial. The interviewer had rescheduled the interview three times, which lowered my expectations from the start. The main focus was on my previous experience and some basic embedded systems trivia, such as defining pull-up and pull-down resistors.

The hiring process involved a demanding five-stage interview that spanned several months. I managed to get through four of the five stages, but then the team's business needs shifted, and they put the hiring on hold indefinitely. The interviews themselves and talking with the hiring team were generally really nice, but the whole process and not knowing what was going on made it super frustrating. I noticed a lot of changes in hiring team members, the role's responsibilities kept changing, and so on, which made it seem like Neuralink has some serious organizational and communication problems internally. It wasn't a total waste of time, and I mainly just gained more experience with technical interviews. The questions were fair, interesting, and directly related to the job, but I really wish I'd gotten direct feedback from the interviewers instead of only hearing from HR.

Got a call from a nice recruiter. Then there was a 1-hour live coding interview. The task was to train a model for their data using PyTorch. It had 10 questions that involved writing or completing code and explaining the answers in writing, most were pretty straightforward. The main difficulty was fitting it all into one hour, which feels a bit unrealistic, but you could get a lot done if you're super fast. You could only use library references, no outside help like stack overflow or autocomplete. The questions felt more like basic data science than ML, using older techniques instead of modern deep learning. It would have been better to have a more challenging, modern deep learning exercise with more time. Overall, it was a good experience, and the interviewer managed the exercise well.