I applied online. The process took 3 weeks. I interviewed at Rakuten in Apr 2017
Interview
Round 1. Record 5 answer to standard interview questions and complete one Codility test with two tasks. The video interview was a nice feature as it could be fit into your schedule as desired. This reduces scheduling issues, particularly when there is a time difference.
Round 2. Complete a presentation, provide a reference letter, and complete two more Codility tasks. Do not bother with the presentation or reference letter until after you do the Codility tests. If you fail the Codility tests and your application will automatically be rejected, then you will feel like an idiot for wasting time on a presentation and wasting someone else's time on the reference letter. Rakuten should split this stage into two: Codilty test then, if successful, presentation and reference letter.
While I understand the need for the first round Codility test, where both tasks were simple enough to be FizzBuzz tests and hence act as a quick filter, the second round Codility test was a debugging problem and a geometry problem (calculate the overlap area of two intersecting circles).
The particular issue with the geometry problem is that it doesn't test coding ability. It tests your willingness to break the honour code and Google the correct answer. Or it penalises the test-taker for not being fresh out of a computer science course where I guess they would be expected to know the answer?
The debugging tasks, heavy focus on Big O metrics and lack of alignment to the applied for job does make me worry that the Rakuten definition of a data scientist is closer to that of a data engineer or more standard computer programmer. Perhaps I dodged a bullet there?
Interview questions [1]
Question 1
(Codility test) Calculate the overlapping area of two intersecting circles
this interview process had 5 rounds
2 was a tech test
３ were conversational in nature
with various managers from direct and indirect to the department manager mostly in Japan but also outside Japan.
1st Interview was simple leetcode task and some questions about Kubernetes. 2nd interview - classical ML and DL questions, Final interview is mostly informal, but there may appear some questions about specific experience and problem solving
Interview questions [1]
Question 1
What is the difference between random forest and boosting?
Initial LeetCode coding test, hiring manager interview for profile and role fit, multiple technical rounds covering statistics, machine learning, SQL, Python, and case studies, concluding with behavioral interviews evaluating teamwork, communication, and cultural alignment.
