Ho presentato la mia candidatura tramite l'università. La procedura ha richiesto 4 settimane. Ho sostenuto un colloquio presso Nationwide nel mese di nov 2013
Colloquio
recerived an e-mail for onsite interview. Met with about 3 people who just chatted about my resume and escorted me out. got an offer in a weeks time. After I got the interview, the HR people contacted me to talk about the other benefits and compensation and terms and conditions. I was asked to pass a drug test.
Domande di colloquio [1]
Domanda 1
What is the guarantee that you will leave the current job?
first sent an email for interview appointment with some questions like expected salary, can you relocate? In the phone interview, ask some qustions like tell me about yourself, why you are applying this job.
Ho presentato la mia candidatura tramite l'università. Ho sostenuto un colloquio presso Nationwide
Colloquio
3 rounds in total:
1: BQ+simple technical (30 mins)
2: take home data challenge, write report(one week)
3: if report passes, present 20 mins to review panel, after this they will ask you follow-up questions
Domande di colloquio [1]
Domanda 1
Do you have experiences with emergent schedule change and how to handle that?
Ho presentato la mia candidatura online. Ho sostenuto un colloquio presso Nationwide
Colloquio
The interview process was very slow. They took around 2 months to make a decision. At first they provide a modeling challenge which you have to solve in a week. You are asked make model and write up a documentation of what was done in detail. Then there is a model review interview(panel of 3-5 people) just like it happens in a office with a MRO.
Ideally you would think that they would have read your documentation & understood your approach when they say you pass but I got the feeling they never read my approach. They just saw the results & since I had great results, they passed me in the initial submission. I made a nice supplementary presentation to show why my model was right but they kept asking questions which were already explain in the submission. (It was pretty clear they did not check the documentation). So if you have a model review next time with them, start with the submission as they would not have checked it mostly.
Also, the dataset they gave is simulated. So I found the best algorithm with great results, That kind of hurt their ego also. Their ideal candidate is someone who will struggle with the data, do lots of feature engineering, spend lot of time & apply lot of science and complex algorithms. When I solved their simulated data with a very simple but mathematically backed algorithm, they kind of had the feeling that I got lucky. But as per me I studied the data, did exploratory analysis & found the best algorithm that fits the model (roc 0.96 classification with low FP and low FN in a skewed dataset).
If they wanted me to behave as I would do with realtime data, they should have provided realtime data. Since they could not accept that, they stopped communicating & sent a reject almost after a month through the automated system.
Domande di colloquio [1]
Domanda 1
They will interlace standard machine learning & data science question as you discuss about the model. Multicolinearity, random forest, logistic regression were some that I can remember.