Are you looking for an NLP Hero to implement your million dollars idea? Here at Almeta we value our employees and our customers alike and we always try to attract top talent. In this article we will give a template for our NLP technical interviews that well hopefully help you in selecting the latest edition to your team.
First start by explaining the parts of this interview e.g. “first we would like to know better about you, next we wish to learn more about your work and educational experience and finally there will be some simple technical questions”
The following are the proposed parts of the interview alongside the upper limit on time for each part and also some possible questions.
Warm up (10 min)
This is meant to be a set of personal and slow questions to get to know the person more.
- Apart from the CV, your academic and professional experiences. Can you tell us more about yourself , like who is … , what are your interests, hobbies, what do you do in your free time?
- Why AI in general and NLP in particular, like why not computer vision?
- What is a specific advancement in AI in general are you interested in?
Candidate dependent questions (20 min)
These will be open-ended questions guided mainly by the candidate’s CV and documents, even a small glance at the social media might be beneficial. Furthermore, they would usually be associated with follow up questions to lead to more detailed technical information and experiences.
Technical questions (10 – 15 min)
It might be a good idea to ask the candidate about a certain problem we have at company and how they would handle it.
- Data cleaning and preprocessing is a very important part of the pipeline, in your preferred language (Arabic, Turkish) what are the steps you would follow to do so? How do you choose
- Specific problem example: (clickbait) explain what is clickbait and ask the candidate how he/she would handle such a task (with regards to the datasets, scraping and modelling)
Situational questions (15-20 min)
Here we present the candidate with a certain situation and ask him/her what they would do in such a case.
Lets assume that you trained a model on a certain task and a certain dataset, and the performance is very good on both validation and test, yet when you try it on production it falls short and below expectation. What are possible reasons behind such a behavior and how would you handle such a task?
Wrapping up (5 min)
Thank the candidate for their time and then answer any questions he/she might have, and ensure him/her that we shall contact them very soon to give them the result.
In this article, we presented a template of a technical ML/NLP interview. This is not a definitive answer and you should always modify this to suit your company needs and goals, but we hope this will help you in choosing the top people in your team as it had helped us.
Do you know that we use all this and other AI technologies in our app? Look at what you’re reading now applied in action. Try our Almeta News app. You can download it from Google Play or Apple’s App Store