Nubank Data Scientist Interview Questions
& Process

The interview process was very long and tiring, with unprepared interviewers. Nubank's process is excessively long (8 stages) and frankly becomes tedious. There's a lot of time between interviews, and I had to chase the recruiter for responses because it was slow. My screening was in early December, and the first technical interview didn't happen until February. The following stages occurred in March, with large gaps (up to a month between technical phases). After the HR screening, there's a CodeSignal test with theoretical ML and algorithms questions (easy level). From there, several technical stages begin. I applied for Senior Data Scientist, but midway through the process, I was simply 'moved' to Machine Learning Engineer (without clear communication). I only found out because I saw the different title on the meeting invitation. If I hadn't asked, I probably wouldn't have been informed. Regarding the interviews: Basic Programming Case (Data Science): simple problems (easy level). The issue isn't the questions, but the interviewers. At one point, I was incorrectly corrected on a basic Python concept ('sorted()' vs. '.sort()') and had to explain to the interviewer that both approaches are valid. And it wasn't a trick question; he genuinely didn't know and said I was doing it wrong. Modeling Case: content was okay (EDA, feature engineering, metrics, etc.), nothing unexpected. MLE Programming Case: the experience here was bad. The prompt was huge, confusing, and the interviewer didn't help clarify doubts. I was basically told to 're-read' when I asked for explanations, which just wasted time. Also, we weren't allowed to read all the questions before starting, and with each implementation attempt, I was interrupted with restrictions that hadn't been explained beforehand ('you can't change this,' 'it's not this way'). It's impossible to code the basics like that. The biggest problem with the process isn't the technical questions, but the lack of preparation from those interviewing. If a person can't explain the problem or answer basic questions, it completely compromises the evaluation and wastes the candidate's time. I've conducted many technical interviews myself and know that clarity and alignment are the minimum expected because time is short. That didn't happen here. I ended up not moving on to the manager stage.

I went through the whole DS interview process but didn't get an offer. It had 5 interviews. First, a HR videocall about my experience and why I want to work at Nubank. Then, Hackerrank questions on data structures and algos. After that, a videocall with a Lead DS about DS questions, like how to impute missing data, the DS project process, and the difference between xgboost and random forest. Then came a Business Case problem, which was the hardest, involving investigating a real-world DS model issue like performance dropping. Finally, an interview with a DS lead about my experience and past DS projects. I'm leaving a negative review because it wasn't clear why I was rejected or what to improve on. Most of my background is as a Data Analyst, so I haven't deployed models at work. I applied for a Data Scientist role, not a senior one, but the interviews focused on models I'd deployed. The HR team was great until the rejection, then they stopped responding, which felt rude. I was told I'd have a final call for feedback after the last interview, but it never happened.

Teve duas fases na entrevista. Primeiro um teste online e depois as entrevistas finais onsite. Nos testes online, havia um monte de questões de múltipla escolha sobre 3 matérias: matemática, estatística e programação. Eram 3 questões de programação onde precisei escrever código nas três.