In my consideration, it would be better to keep the company’s name hidden and I will extend with the assignment including the questions and my solution after the position closed or later.
After I finished my master mid-term defense, I spilled my resume to thousands of internship recruitment on the Linkedin and finally, there’s one sent me an invitation for the interview (probably the only one lol).
Major: EE (electronics engineering) based, CS (computer science) for half an year, more focusing on modeling deep neural network. (Yup, you are right, that's exactly a newbie's excuse)
Research: Arrhythmias detection, Multi-class task, Deep learning, Autoencoder.
Position: Data Scientist Intern (Japan)
JD: Be able to use statistical modeling, supervised & unsupervised learning and clustering (more fancy to use prediction model and clustering method).
There are two parts involved in this section, one is my research introduction and another is a quiz for me, and we spent about one hour for all.
For the research introduction, it is about the time-controlling. If you understand every detail over your research, then you could bring your interviewer into it easily. Once interviewer received and understood well, that is successful leading you to earn more time for next part — the quiz.
In the second part, interviewer put forward an open quiz, no target, open as wide as the skyline. The attribute of data is time-series, the information contains address and time being occupied by the customers (I forgot the definite question and context, but it's similar to those on Kaggle I think). Based on the attribute and the information of the dataset, in my opinion, the interviewer wants to know the overall scope you made is critical and logical enough. And I screwed it up because I haven't an experience of dealing a time-series dataset (Read the JD carefully!!).
At the end of the interview I got an assignment and required to finish in 4 hours, also submit it by next week (And I just push the deadline...). The assignment contains of two questions, the first is to build a predicting model and the second is about the clustering and labeling. Fortunately, I haven't any base on those so I must to learn it forcedly lol.
About one week later I submitted, the HR sent me the mail for the second interview for the follow-up conversation of the assignment and my solution.
The purpose of this second interview is having discussion about the submitted solution which spent about one hour this time too. This time, there is one more guy attended.
The weakness is that I didn't set all up of the explanation for my solution even if my model is simple but reach the performance beyond our expectation.
The best strategy on preparing the second interview is to prepare as the master defense. They will challenge your opinion like your panelists, for instance, why and how you make the decision if you do data reduction and what supports you to do it? (This is the bloody real question but I didn't answer well...) Noted that treat every assignment, homework, coming quiz from the interview carefully.
One hour later, the whole process this time is poor, poor indeed (As the scene Tim meets Mary in the Kate Moss’s exhibition). Now, all I can recall is a disaster like a hurricane destroying and shouting in my brain, but nothing alive.
I have to confess that I underestimate the second time. It is sad but luckily the meeting reflects my weakness and lights out the space where I could keep move forward. Although I have not receive the outcome because the second interview just ended today.