For today’s data scientists, technical skills are necessary but not sufficient.

Photo by "My Life Through A Lens" on Unsplash

A recent report carried out by Ipsos MORI into the UK Artificial Intelligence (AI) labour market supports an often overlooked fact in the industry — technical skills are necessary but not sufficient for today’s jobs in AI.

Of course technical skills are necessary. You’ll need to understand AI concepts and algorithms, have experience with programming (Python being the primary data science language today), be familiar with current AI-related software offerings (cloud AI offerings and infrastructure, open source tools, etc.), be adept at data wrangling, have a firm grasp of mathematics and statistics, etc.

But is that enough?

In my experience — and supported by the aforementioned study — the answer is No.

Let’s take a look at some of the questions I ask when interviewing prospective data scientists.

Technical

• Give me a 1 minute run down of your background and how you evolved into the role you’re in today?
• Can you tell me the difference between AI/ML/DL and which term do you use the most?
• Tell me about your Machine Learning background.
• What’s the hardest part of Machine Learning?
• Tell me about your experience with AWS/Azure/GCP, especially in terms of BigData, Analytics, and Machine Learning.

These questions let me explore the candidate’s background and understanding as well as make sure their resume is legitimate (you’d be surprised how much people exaggerate or down right lie on their resume!). If there are any areas of technical concern, this is where we can dive deep.

Soft Skills

• How would you explain Machine Learning to someone with no experience?
• Give me a pitch to make me excited about using Machine Learning?
• How would you suggest someone gets started with Machine Learning?
• Are you comfortable talking to crowds and guiding workshops?

Public speaking, breaking down advanced concepts into understandable terms, deeply understanding the value proposition (and effectively delivering it), understanding the learning process — these are all key aspects of a role in AI today.

Sure, in the past, there were those techy folks that were great at programming or the like, but you dared not let them speak to customers or management due to their less-than-stellar social skills. But in today’s world, especially with emerging technologies like cloud, AI, blockchain, quantum computing, etc., most people don’t understand the technology and so your ability to effectively explain things and raise excitement is critical.

You can have amazing technical chops, but if you can’t pitch a proposal that is understandable and exciting, you most likely won’t get the chance to display those chops. Understanding how to build neural networks is important, but if you can’t grasp the larger context of the problem from a business-perspective, then what’s the point? If you’re an academic researcher, then go ahead and solve some novel problem for the sake of science. In the real-world, AI is simply a tool (just one of many) used to solve a problem.

Humanity

• What is the coolest thing you have learned on your own that has helped you better perform your job?
• Give me an example of when you have had a genuine interest in someone other than yourself.

At the end of the day, the people we work with, as well as our customers, are just people. And people want to do business with people they like. Give me an average data scientist that’s a really cool person over a megalomaniac rock star any day. How do you talk about your interests to others? What is it that interests you (and why)? How do you relate to other people? These are all extremely important aspects in a coworker. Laziness, drama, self-centeredness, toxicity — these are things to avoid at all costs.

Preparation and Curiosity

• Do you have any questions for me?

I try to have fully half of our conversation (yes, an interview is a conversation to be valued, not an interrogation to be feared) on this one question alone. What the candidate asks is almost always far more insightful than anything I ask them. What matters to them? What have their previous experiences been like? These will often drive the questions they ask. What are their preparation skills? If you have no questions for me at this point, you probably won’t get the job.

Of course, there are good, but fairly standard, questions to ask the interviewer.

• What is your ideal candidate?
• What is most important for this role?
• How would you explain the culture of the company/team?
• What does a typical day look like?

And, of course, there are bad questions to ask as well (unless you’re in final negotiations where details matter).

• What would be my exact hours?
• How much vacation time would I receive each year?
• I would never be expected to work over 40 hours, would I?
• What is the salary for this role?
• Which specific software tools do you use?

There are also more insightful questions.

• I want to be part of a team. What is the team structure? Who would I be working with and what are they like?
• I prefer a faster pace. How would you classify the pace of this role?
• How do you feel the interview went? What would you say were my strengths and weaknesses as well as areas that could be improvement?

The first two questions are examples of the candidate knowing what they want and investigating if this exists within this particular role. This can be extremely powerful if asked correctly.

The last question (or some variation of it) is one that every single candidate should always ask at every interview. It shows your vulnerability as a human being, your capacity to accept criticism, your desire to better yourself, your curiosity. These are all great things to showcase. What do you get out of it? Your interviewing skills will dramatically improve from the feedback. You won’t sit for days/weeks wondering how you did or if you will get a call or not. Sadly, of the hundreds of candidates I’ve interviewed through the years, only three have asked this question — and I remember each one of them.

Wanna go the extra mile and make an impression? Be creative with what you ask!
A recent example was when I asked a candidate the “coolest thing you have learned on your own” question. Their answer (at least the first two sentences of it) was: “Overall theory of Bayesian statistics. It’s a philosophy that there may be a core truth but the issue isn’t finding the right answer but getting closer and knowing how close you are.” I found this answer interesting. Then when it came time for his questions for me, the first one out of his mouth was: “What is your Bayesian statistics [meaning what excites me, the interviewer?] and how much does that fit into your day to day?” Weaving together separate parts of our conversation, showing interest in me as a human being, asking about the day to day job in a unique way. I found this question interesting on several levels. He got the job.

Back to the topic at hand

As you can see, the majority of what I focus on during an interview is non-technical, but rather soft-skill based. Effective communication, understanding the bigger picture, commercial awareness, curiosity, being a decent human being — these are the things I tend to focus on.

Of course you need some technical proficiency — I mean, if you’ve never heard of the internet, you’ve probably got a pretty steep uphill climb. But as long as you have the baseline technical capabilities, I can teach you the rest. It’s far easier to teach someone about XGBoost or Semantic Segmentation than it is to teach them to be curious, or an effective communicator, or not a jerk.

To succeed in this role today, you need to be a developer as well as a salesperson.

University and on-the-job-training often teach people technical skills, but these alone are insufficient to really succeed in the world of AI. Were I to develop a curriculum for a degree in AI, I would include classes on mathematics, statistics, engineering, programming, data wrangling, cloud infrastructure, and machine learning (frameworks, algorithms, hyperparameter optimization, model drift, etc.) — all things you would expect. But I would also include classes in business, ethics, philosophy, speech, communication, critical thinking, and leadership.

“If I had an hour to solve a problem I’d spend 55 minutes thinking about the problem and 5 minutes thinking about solutions.”
― Albert Einstein

So, the next time you look at your resume, look beyond just “I utilized PCA for feature engineering” or “I developed a model using the BlazingText algorithm”. Demonstrate your understanding of the business problem you were solving and how these AI techniques helped. What was the ROI? Consider using the STAR (situation, task, action and result) method to communicate your thoughts clearly.

The next time you’re in an interview, relax and enjoy the two-way communication. Be authentic, curious, thoughtful and interesting. Are you coming off as the type of person someone would want to work with? Does the interviewer seem like someone you would like to work with? Remember, the company is interviewing you but you are also interviewing the company. Too many people don’t seem to understand this.

When you’re on the job, remember that it is more important to be able to explain things in simple words to a client or team member than it is to increase the accuracy of your model from 90% to 93%. Be on the lookout for opportunities to utilize (or enhance) your leadership capabilities. Be curious not only about how to solve a problem with AI but also about the privacy and ethical issues surrounding AI. Understanding the big picture of the business problem we are trying to solve should be your number one goal. Without this understanding, you may very well use your technical skills to solve the wrong problem.