The world’s leading publication for data science, AI, and ML professionals.

What 50+ ML Interviews (as an Interviewer) Have Taught Me

What can you do as an interviewer to leave a positive impression on a candidate regardless of the outcome?

If you thought facing a technical interview is hard, try conducting an interview. I’m not talking about the awkward interviewers who left a scathing impression, watching you condescendingly, while you’re wishing for the pain to end. I’m talking about interviewers that have left a positive impression you cherish.

Doing an interview is a huge responsibility. You are the gatekeeper for someone’s career. So you must do everything in your capacity to do them as much justice as you can.

All of the points I’m going to be mentioning revolve around one key value:

Empathy!

If you’ve mastered empathy, you probably don’t even need to read this. Everything I talk about here are my personal thoughts and opinions and do not reflect the view of my employer.

For context, I’m a machine learning engineer and these are technical or coding interviews I’m talking about.

After reading this article, you’ll (hopefully) take away a few lessons that will make you a better interviewer, who leaves a positive lasting impression on the candidates (regardless of the outcome).

Let’s skip the obvious checks for a candidate and clear the air first:

  • Showing sound technical knowledge of the language & tools
  • Thinking out loud while working through the problem
  • Being friendly and cooperative

Any of these that don’t get a tick is a concern.

Collaboration over scrutiny

The style in which interviewers conduct the interview plays a massive role on how an interview plays out. If the interviewer appears as an authoritative figure that cannot be pleased in anyway, even the best candidates may fold and get nervous. The candidates are at their best when they have someone approachable in front of them.

One way I do this by is by making the interview almost a pair programming session, in which two of us are trying to get to the solution, rather than silently (somewhat sadistically) watching every move of the candidate while they’re grimacing at the question.

I won’t lie, I used to root for "team scrutiny". At the time, I believed the most important thing to evaluate the candidate on is to check if they can find to the solution on their own. On the job, time-to-solution seldom depends solely on your technical knowledge. Rather, it’s dictated more by factors like…

  • Knowing the right questions to ask (from right people of course)
  • Communicating your work so far
  • Interpreting and executing someone’s hints/suggestions

Pair programming gives you a golden opportunity to test the candidate on these "soft-er" skills. So rather than a solo effort, the pair embark on a journey, harmoniously working to get to the final solution. To nudge the candidate in the right direction or to break the ice when stuck, I’d use phrases like "tell me what you’re thinking" or "maybe start with pseudo-code". I’m not saying you should hand over the solution to them, but helping them to succeed.

Not only this approach helps you to evaluate the candidate on a range of hard and soft skills, this leaves the candidate with a sense of accomplishment (as opposed to leaving them out to dry, should they get stuck).

Situational awareness

A key skill you must develop is not remembering the questions by heart (although that’s important), but having an acute situational awareness. Every candidate is different, and every interview experience is different. As an interviewer you must be able to play to the tune of that unique scenario.

If the candidate appears to be nervous, smile a bit more, ask about the weather and let them know you’re here to help if needed. If the candidate needs some personal space when solving a problem, give them that. If the candidate had technical difficulties at the start of the interview, give them a bit of extra time to make up for the lost time.

Not only you have to do this during the interview, but afterwards as well. When evaluating the candidate, build a fair image of their performance by cutting through noise. Ask yourself questions like,

  • Could they have done a better job if they weren’t nervous?
  • Did the candidate have a cultural/language-barrier?
  • Is this someone I’d look forward to working with?

The candidates typically get more and more confident/comfortable as they go through the stages of interviews. Perhaps the candidate is having a really bad day! So as an interviewer, you need to be able to cut through these and evaluate the candidate on their skills. I’m not saying you should completely ignore these facts. Note them and communicate them to the other interviewers/recruiters. But if you expect an interviewee to be at their prime always, you’re in for some disappointment. Remember we’re all human! And this gives us a nice Segway to the second point.

Write detailed feedback

This is one of the most difficult things to master as an interviewer: writing up a comprehensive and fair evaluation. Why is it difficult? Because there’s many things to evaluate the candidate on within an hour or a 45-minute window with your limited mental bandwidth. As with other things, it takes time to master this skill.

Few things that can help you are,

  • Use a template to take notes during the interview. It could be something like, for each task in the interview, what they did well and what they didn’t do so well. One important thing is have concrete examples than vague statements.
  • Go through the feedback others have written (if you have visibility over that)
  • Don’t put off taking notes – make sure your notes are complete by the time you end the interview. Otherwise, you’re ALWAYS going to forget the important observations you made during the interview.

Be punctual

Being at the other end of the table, it’s easy to summon a sense of royalty. Thoughts like, "they should be waiting for me to enter, not the other way around" may creep in. Personally, it’s more important that I value someone else’s time the same way they (or I) do mine.

I don’t have to remind how exciting (in a horrifying way) it can be just before the interview. And 99% of those candidates would be sure to join the interview on the dot. It’s only fair that I do the same for the candidate as an interviewer.

Conclusion

Conducting a great interview is not a science. There’s no magical recipe that’ll work always. Rather it’s based on your ability to read the situation and the candidate.

It’s important to take these advice/learnings with a grain of salt. They are not exact rules, nor there are exact measurements. For example, I’m not saying you should pass a candidate only because they were nervous. But give them a second chance if they got up to 80% of the answer and you know that it’s been a while for them and that they were quite nervous.

Additionally, you being a "good human" can help to make the interview a pleasant experience. Sadly, I’ve seen interviewers mocking weaknesses of candidates afterwards. An interview for most of us is a really stressful experience and we all have made silly mistakes at some point. So try not to do that.

Thank you for reading! I hope these tips would help you to become a better technical interviewer!

If you enjoyed this story, feel free subscribe to Medium, and you will get notifications to fresh content from me, as well as unlock full access to thousands of engaging stories from other authors.

Join Medium with my referral link – Thushan Ganegedara


Related Articles