The process was rigorous and tedious. You're given a dataset (multiple tables) and an open ended questions. The recruiters says that it takes about 4 hours to do, but generously provided a week to get the project submitted. It turned out that it indeed took a week to get it done as you'll have to built things from scratch. The interview process was more like university exam: tell me step by step about gradient decent algorithm. Explain how CNN works. What's the formula for L1 regularization, etc. There are about 10 similar text book or formula memorizations questions. They are pretty easy but can get you caught off-guard since they are not practical. The hard part was interviewing with the manager --I was surprised how moody they are. My interview was at 9 AM, online, and the manager was in her bedroom, just got up from bed with hairs all over the places and cranky (presumably due to lack of coffee).
Interview questions [2]
Question 1
Tell me the formula about L1 and L2 regularization.
Thanks very much for taking the time to leave a review, and we're sorry that your interview experience wasn't more enjoyable.
At this time when we are all adjusting to a new working normal, we are all working from home, and some of us from our bedrooms. We'll be sure to pass your feedback on, but we'd also ask that you fill in the survey that was sent out to you.
I applied online. The process took 6 weeks. I interviewed at Revolut (Remote, OR) in Mar 2020
Interview
The process took somewhat 4-6 weeks. I applied for the job position located in Poland. The recruitment process was fully remote and consisted of hangout calls. The first step is the CV screening and a call with the HR guy. The next step is a homework assignment, consisting of 2 exercises (one, is an SQL problem - a redshift-powered metabase, second, is a machine learning challenge). The third step requires you to: a) present your ML solution; b) answer some theoretical DS questions; c) pass Python live coding exercise; d) again, answer some DS questions. The fourth step is the final interview with the head of the AI department.
Interview questions [8]
Question 1
How to exactly implement the calculation of a median using a map-reduce algorithm? It must work well with hundreds of millions of numbers. Describe all steps of the algorithm. What approximation methods of a median you know?
You trained recently a gradient boosting model. You need to retrain it from time to time, also, it must be put to production. There are no data engineers nor ml-ops people to help you. What would you do?
I applied in-person. The process took 5 weeks. I interviewed at Revolut (Londen, Engeland) in Feb 2020
Interview
The interview process took slightly more than 5 weeks. Divided into 4 main stages, the process is extremely efficient thanks to the DS talent team. At each single stage, thorough feedback was provided.
The four stages consist of:
- Phone interview with talent lead (general questions about yourself and some preliminary technical questions around Python, SQL)
- Take home data challenge (quite long to finish but extremely interesting). One week time is available for the candidate to complete it
- Technical interview with 2 senior data scientists. Here you are asked to present findings of the data challenge, answer technical questions on DS or machine learning, do a live python coding exercise
- Final interview with head of team or pm: this is mainly a culture fit assessment as well as previous background and experience check to make sure team's goals and candidate skills are aligned.
Interview questions [3]
Question 1
Questions around Bayesian methods used in data science